1
Sede Amministrativa: Università degli Studi di Padova
CISAS - Centro d’Ateneo di Studi e Attività Spaziali ―G. Colombo‖
___________________________________________________________________
CORSO DI DOTTORATO DI RICERCA IN: Scienze Tecnologie e Misure Spaziali
CURRICOLO: Misure Meccaniche per l'Ingegneria e lo Spazio (MMIS)
CICLO XXIX
Blade Tip Timing and Blade Tip Clearance
Measurement System Based on
Magnetoresistive Sensors
Coordinatore Corso: Ch.mo Prof. Giampiero Naletto
Coordinatore Curricolo: Ch.mo Prof. Stefano Debei
Supervisore: Ch.mo Prof. Gianluca Rossi
Dottorando: Roberto Tomassini
2
3
"Innovazione è vedere ciò che hanno visto tutti
e pensare ciò che nessuno ha pensato"
Albert Szint Gurgyi
4
5
Abstract
The accurate control of running gaps between static and rotating components is vital to preserve
mechanical integrity and ensure a correct functioning of any rotating machinery. In aero-engines,
tip leakage above the airfoil tip results in relevant aerodynamic losses, therefore the gap between
the rotor blades tip and the casing must be the smallest possible, but enough to avoid contacts. One
way to measure and monitoring tip gap is by a Blade Tip Clearance (BTC) technique.
The measurement of blade vibrations and eigen frequencies is also a crucial part of the development
of any new gas turbine or aero-engine. For more than half a century, this has been performed by
installing strain gauges on the blades and using telemetry to transmit the signals. The Blade Tip
Timing (BTT) technique, (i.e. measuring the blade time of arrival from the casing at different
angular locations with proximity sensors) is currently being adopted by all manufacturers as a
replacement for the classical strain gauge technique because of its non-intrusive character.
However, the aero engines presents an extremely challenging environment for the application of
reliable in-service instrumentation. Current studies focus on sensors development, new data
processing methods and the realization of a standard for such measurements.
In this thesis, a simultaneous BTC & BTT measurement system was designed, realized and
optimized using magnetoresistive sensing elements. A simultaneous BTC & BTT system ensures
the measurement of the gap between the rotor blades and the casing and the measurement of blade
vibrations. Different types of probes have been proposed and are currently used to monitor these
parameters: optical, capacitive, microwave, eddy current. The sensors developed in this research
project are based on the magnetoresistivity. Dedicated hardware and software components had to be
designed and manufactured. Two external test at Fogale Nanotech in France and at the Military
Research Institute of Technology (ITWL) in Poland are reported. The measurement chain was fully
characterized and the innovative instrumentation for BTT and BTC measurements based on the
magnetoresistivity has achieved promising performance for a professional product.
This work was realized in a collaboration between the ―Centro di Ateneo di Studi e Attività Spaziali
G.Colombo‖ – CISAS, Padova, Italy, the Engineering Department of the University of Perugia,
Italy and the Von Karman Institute, Rhode Saint Genese, Belgium.
6
Summary
During the development of a new machine there are two fundamental aspects to take into account:
the mechanical structure of the machine and its efficiency. A fundamental part of the design and the
development of a rotating machinery is the determination of reliable and performing geometries.
From a mechanical and structural point of view, the machine components must provide a certain life
in safe conditions, for instance avoiding resonance phenomena in the working range of the machine.
At the same time, the machine components must provide also an high efficiency. It is not the case
that the perfect structural part it is also the most efficient, very often the definitive geometries of a
turbomachine are a compromise between structural and efficiency requirements. In this scenario it
is necessary to have a robust design process and numerical analysis of stress and strain levels, of the
loads as well as the performance of the parts of the machine. Finally, it will be mandatory to have
experimental validation sessions. Considering for instance an industrial gas turbine or aero-engines,
it is fundamental that the rotor blades and disks will be accurately designed for a long life (fatigue
life), safe working conditions far away from resonances and dangerous fluid-structure interactions
and characterized by high efficiency. The design team and the aerothermal team must collaborate in
designing and defining geometries that, numerically, will be reliable and that provide high energetic
machine efficiency. Fundamental will be than the experimental validation of the prototypes, carried
out in test benches properly equipped. This last part is realized using professional instrumentation to
validate the prototypes and to certify their performance. In particular, concerning rotor blades and
disks, the experimental test regards the monitoring and measurement of blade vibrations during
operation and the running gaps between the blade tip and the casing, key parameter for an high
aerodynamic efficiency. The instrumentation used can be with contact with the rotating parts or
contactless. The latter are the so-called Blade Tip Timing (BTT) and Blade Tip Clearance (BTC)
measurement systems.
The aim of this PhD thesis is the development of a BTT & BTC measurement chain based on new
magnetoresistive sensors. This instrumentation is already commercially available and it is used in
turbomachinery testing since years, but there is a continuous development in the frame of new
sensor technologies and data processing methods.
During the first year of the PhD new magnetoresistive probes were designed and realized to provide
simultaneous measurements of ―blade tip gap‖ and ―blade vibrations‖. A dedicated electronic and
software for the signal analysis had to be developed. The work started from preliminary test using
the magnetoresistive sensing elements and carring-out the characterization at the passage of
7
ferromagnetic parts. Then, test benches were developed to deeply study the layout of the sensor and
to analyse the signal to noise ratio at the blade passage. Once the probe geometry was defined, the
first year ended with the realization of a dedicated electronic for the signal amplification and
filtering and a software for the signal analysis.
The second year of the PhD was focused on the experimental characterization of the measurement
chain and on the development of mathematical models. Once the instrumentation was completed, it
was the time to characterize fundamental parameters like the rise time, uncertainty levels on blade
tip gap measurements, definition of key parameter for the vibration measurements and the studies
on the influence of external sources of errors. To extend the know-how, 2D and 3D mathematical
models were realized to simulate operating conditions and for future developments.
The last part of this research work was the validation of the instrumentation, the optimization of
hardware and software and comparisons with existing BTT and BTC instrumentation. In particular,
once the whole measurement chain was realized and characterized, it was necessary to validate the
instrumentation against some commercial and certified probes. For that reason, external test were
performed in the test benches of Fogale Nanotech in France and in the second compressor stage of a
real military jet-engine at the military research institute of technology (ITWL) in Poland. The
results were very satisfactory and it was possible to improve the performance and the accuracy of
the whole system.
8
Sommario
Nello sviluppo di un macchinario ci sono due aspetti fondamentali che devono sempre essere
considerati: la meccanica della struttura e il suo rendimento. Una parte fondamentale della
progettazione e sviluppo di una macchina rotante è la definizione di geometrie che siano affidabili e
performanti allo stesso tempo. Da un punto di vista meccanico-strutturale, i componenti in
movimento devono garantire una certa durata nel tempo rispettando dei margini di sicurezza, come
operare in un regime lontano da risonanze e fenomeni dannosi. Se da un lato una certa geometria
può soddisfare tali requisiti, dal punto di vista energetico il macchinario dovrà essere efficiente.
Non è detto che il componente strutturalmente perfetto sia anche efficiente, anzi, molto spesso nella
progettazione di una nuova turbomacchina, la versione definitiva prevede un compromesso fra i due
aspetti. Tali esigenze richiedono un rigoroso metodo di progettazione e analisi numerica degli stati
di tensione, delle sollecitazione e delle performance delle componenti del macchinario che
dovranno poi, imprescindibilmente, essere validate sperimentalmente. Considerando ad esempio il
caso di turbine a gas industriali o motori aeronautici, è fondamentale che i componenti della
turbomacchina e, in particolare palette e dischi rotanti, siano correttamente dimensionati per una
lunga vita a fatica, che in condizioni di lavoro si trovino ben lontani da risonanze e fenomeni di
interazione fluido-struttura potenzialmente dannosi e che presentino alti rendimenti. Il team di
progettisti strutturali insieme al team degli esperti di fluidodinamica dovranno così dimensionare e
definire delle geometrie che, numericamente, siano strutturalmente affidabili e allo stesso tempo
garantiscano un alto rendimento energetico della macchina. Fondamentale è poi la validazione
sperimentale dei prototipi proposti, fatta su banchi prova opportunamente strumentati. Questa
ultima parte è condotta utilizzando strumentazione professionale fine a validare i prototipi e
certificarne le performance. In particolare, sempre con riferimento a palette e dischi rotanti,
sperimentalmente vengono monitorate e misurate le vibrazioni delle pale in esercizio e il gap
esistente tra la punta della pala e il casing, parametro chiave per un’alta efficienza aerodinamica. La
strumentazione utilizzata può essere a contatto con le parti in moto, o di tipo non a contatto. In
questo ultimo caso si parla di sistemi di misura Blade Tip Timing (BTT) e Blade Tip Clearance
(BTC).
L’argomento di questa tesi di dottorato è lo sviluppo di una catena di misura BTT & BTC basata su
nuove sonde magnetoresistive. Tale strumentazione già esiste ed è ampliamente usata nel settore
delle turbomacchine, ma è in continuo aggiornamento sotto gli aspetti dello sviluppo di nuovi
sensori e nuove tecniche di analisi e processing dei dati.
9
Durante il primo anno del corso di dottorato sono state progettate e realizzate le sonde analogiche
magnetoresistive per misure simultanee di ―blade tip gap‖ e ―blade vibrations‖. Elettronica dedicata
e software di analisi del segnale sono stati sviluppati di conseguenza. Si è partiti da prove
preliminari con gli elementi sensibili magnetoresistivi e loro caratterizzazione al passaggio di
componenti ferromagnetiche. Si è poi sviluppato dei banchi prova per uno studio approfondito del
layout del sensore e l’analisi del rapporto segnale rumore al passaggio di palette di turbomacchine.
Definita la geometria della sonda e i suoi componenti, il primo anno si è concluso con la
realizzazione dell’elettronica dedicata per l’amplificazione e filtraggio del segnale e il software per
le misure a partire dall’analisi del segnale.
Il secondo anno di dottorato si è concentrato nella caratterizzazione sperimentale della catena di
misura e nello sviluppo di modelli matematici. Completata la realizzazione della strumentazione,
sono stati caratterizzati i parametri fondamentali quali rise time, incertezza in misure di blade tip
gap, definizione dei parametri chiave per il calcolo delle vibrazioni e studio dell’influenza di
sorgenti esterne di disturbo. Per completare il know-how sviluppato sono stati realizzati dei modelli
matematici 2D e 3D per la simulazione delle prove sperimentali e per utilizzo in sviluppi futuri.
L’ultima parte del dottorato si è focalizzata sulla validazione della strumentazione, l’ottimizzazione
delle componenti hardware e software e comparative con strumentazione esistente. In particolare,
realizzata e caratterizzata l’intera catena di misura, è stato necessario validare la strumentazione con
delle sonde commerciali professionali certificate. A tal fine, è stato possibile eseguire dei test su
banco prova presso Fogale Nanotech, in Francia e sul secondo stadio di compressore di un motore
aeronautico militare presso l’istituto di ricerca militare (ITWL) in Polonia. Risultati più che
soddisfacenti hanno confermato la validità dei componenti sviluppati e dei modelli realizzati,
consentendo di migliorare le performance generali e l’accuratezza di misura.
10
Contents
Abstract ................................................................................................................................................ 5
Summary .............................................................................................................................................. 6
Sommario ............................................................................................................................................. 8
List of figures ..................................................................................................................................... 14
1 Introduction ................................................................................................................................. 19
1.1 Background ......................................................................................................................... 19
1.2 Literature review ................................................................................................................. 20
1.2.1 BTC and BTT measurement systems........................................................................... 20
1.2.2 Measurements without holes ........................................................................................ 21
1.2.3 Blade Tip Timing Data Processing .............................................................................. 21
1.3 Objective of the research project ......................................................................................... 23
1.4 Structure of the thesis .......................................................................................................... 23
2 Fundamentals of blade vibration and blade tip clearance problems in turbomachinery ............ 24
2.1 Introduction ......................................................................................................................... 24
2.2 Blade modes ........................................................................................................................ 24
2.2.1 Classification of blade modes ...................................................................................... 24
2.2.2 Parameters affecting the blade mode properties .......................................................... 29
2.3 Disk modes .......................................................................................................................... 32
2.4 Bladed disk modes............................................................................................................... 36
2.5 Bladed disk response ........................................................................................................... 42
2.5.1 Engine order excitations............................................................................................... 42
2.5.2 Response of a bladed disk ............................................................................................ 42
2.5.3 Classification of bladed disk response ......................................................................... 46
2.5.4 Single and double modes ............................................................................................. 49
2.6 Campbell diagram ............................................................................................................... 50
11
2.7 Interference diagram............................................................................................................ 52
2.8 Blade tip clearance variations.............................................................................................. 56
2.9 Conclusions ......................................................................................................................... 57
3 Blade Tip Clearance and Blade Tip Timing measurement systems ........................................... 58
3.1 Introduction ......................................................................................................................... 58
3.2 The principle for the measurement of blade tip clearance .................................................. 58
3.3 The principle for the measurement of blade vibrations ....................................................... 60
3.4 Conclusions ......................................................................................................................... 63
4 Design of the hardware and the software of the measurement system ....................................... 64
4.1 Introduction ......................................................................................................................... 64
4.2 The magnetoresistive sensing element ................................................................................ 64
4.3 Design of the sensor ............................................................................................................ 67
4.4 The first sensor prototype .................................................................................................... 70
4.5 The mathematical model of the sensor ................................................................................ 70
4.6 The basic electronic board ................................................................................................... 72
4.7 The Universal Calibration Bench for BTT & BTC systems ............................................... 73
4.8 Design of the post-processing software............................................................................... 76
4.9 Conclusions ......................................................................................................................... 82
5 Blade tip clearance measurements .............................................................................................. 83
5.1 Introduction ......................................................................................................................... 83
5.2 Characterization of the magnetoresistive sensor at the blade passage ................................ 83
5.2.1 Test setup and methodology ........................................................................................ 83
5.2.2 Results .......................................................................................................................... 83
5.3 Preliminary clearance test with a gear ................................................................................. 84
5.3.1 Test setup and methodology ........................................................................................ 84
5.3.2 Results .......................................................................................................................... 85
12
5.4 Clearance test in the VKI R2 compressor rig ...................................................................... 86
5.4.1 Test setup and methodology ........................................................................................ 86
5.4.2 The calibration curve of the magnetoresistive sensor for clearance measurements .... 87
5.4.3 Uncertainty analysis ..................................................................................................... 89
5.4.4 Results .......................................................................................................................... 89
5.5 Conclusions ......................................................................................................................... 90
6 Blade tip timing measurements .................................................................................................. 91
6.1 Introduction ......................................................................................................................... 91
6.2 Simulations of synchronous and asynchronous vibrations.................................................. 91
6.2.1 Test setup and methodology ........................................................................................ 91
6.2.2 Results .......................................................................................................................... 92
6.3 External validation test at Fogale Nanotech ........................................................................ 94
6.3.1 Test setup and methodology ........................................................................................ 94
6.3.2 Results .......................................................................................................................... 95
6.4 External validation test at ITWL ......................................................................................... 98
6.4.1 Test setup and methodology ........................................................................................ 98
6.4.2 Results .......................................................................................................................... 98
6.5 The VKI R2 compressor.................................................................................................... 101
6.5.1 Test setup and methodology ...................................................................................... 101
6.5.2 Results ........................................................................................................................ 102
6.6 Conclusions ....................................................................................................................... 103
7 Developments ........................................................................................................................... 104
7.1 Introduction ....................................................................................................................... 104
7.2 The real-time software....................................................................................................... 104
7.2.1 Tip-Timing and Tip Clearance data acquisition in streaming .................................... 104
7.3 The differential sensor ....................................................................................................... 107
13
7.4 The Anti-Aliasing algorithm for tip-timing data ............................................................... 108
7.5 Conclusions ....................................................................................................................... 111
8 Conclusions and Recommendations for Future Work .............................................................. 112
8.1 Conclusions ....................................................................................................................... 112
8.2 Future Work ...................................................................................................................... 112
9 References ................................................................................................................................. 114
Acknowledgements .......................................................................................................................... 121
Appendix .......................................................................................................................................... 122
A. The VKI R2 test rig ..................................................................................................................... 122
B. The magnetoresistive sensing element - Honeywell HMC1501 ................................................. 128
C. The data acquisition system ........................................................................................................ 132
D. Blade Tip Clearance .................................................................................................................... 133
E. The Universal Calibration Bench ................................................................................................ 136
F. Optimization of the sensor components....................................................................................... 138
14
List of figures
Figure 2.1 First 3 bending modes of beam with one end fixed and one end free .............................. 25
Figure 2.2 First 3 bending modes of beam with both ends fixed ....................................................... 25
Figure 2.3 Mode shapes of cantilevered thin shells a) First flap wise bending mode; b) First
torsional mode ; c) First edgewise bending mode .............................................................................. 26
Figure 2.4 Nodal lines of modes shapes from cantilevered blades .................................................... 26
Figure 2.5 Mode shape classification of a cantilevered blade ........................................................... 27
Figure 2.6 Non-vibrating radial compressor blade ............................................................................ 27
Figure 2.7 Two vibrating modes of a radial compressor blade .......................................................... 28
Figure 2.8 Two vibrating modes of a radial compressor blade with nodal line indication ................ 28
Figure 2.9 Taper of blades in a) thickness and b) chord .................................................................... 29
Figure 2.10 Equivalent stress in blade outlet part at hub due to centrifugal forces ........................... 30
Figure 2.11 First bending frequency of a blade as function of the blade impeller ............................ 30
Figure 2.12 Centrifugal stiffening: a) tangential vibration; b) axial vibration ................................... 32
Figure 2.13 Deformation of disk with a) only ND(1 NC) and b) only NC ........................................ 33
Figure 2.14 Axial deformations of a disk with m = number of ND and n = number of NC ............. 34
Figure 2.15 Disk vibrations ............................................................................................................... 34
Figure 2.16 Families of a disk mode .................................................................................................. 35
Figure 2.17 Presentation of 4 ND mode with displacements in axial direction ................................. 35
Figure 2.18 First bending modes of blades coupled through a disk: a) in-phase vibration; b) counter
current vibration; c) one blade at rest ................................................................................................. 36
Figure 2.19 Indistinguishable 4th and 32nd nodal diameter pattern of a 36 bladed disk .................. 37
Figure 2.20 Disk stiffening effect caused by increasing number of nodal diameters ........................ 38
Figure 2.21 Effect of disk stiffeness on the bladed disk modes ......................................................... 38
Figure 2.22 Nodal diameters in bladed disk with 6 blades in each disk sector ................................. 39
Figure 2.23 Mode shapes of one disk sector where a black arrow denotes the position of a nodal
diameter .............................................................................................................................................. 40
15
Figure 2.24 Blade motions during vibration: a) No vibration, b) 1ND with translation along
diameter, c) 0ND with torsional disk movement ............................................................................... 41
Figure 2.25 Sinusoidal distortioned flow due to the stator vanes (upper graph) and the sinusoidal
representation of the rotor blades (lower graph) ................................................................................ 43
Figure 2.26 Nodal diameters and excitation of the same and not the same EO ................................. 48
Figure 2.27 Nodal diameters of orthogonal standing waves of a 12 bladed disk .............................. 49
Figure 2.28 Campbell diagram of blade on rotating disk .................................................................. 51
Figure 2.29 Campbell diagram for the first blade flapwise bending mode shape (at all possible nodal
diameters) of a 36 bladed disk ........................................................................................................... 53
Figure 2.30 Campbell diagram for 4 bladed disk mode shapes (at all possible nodal diameters) of a
36 bladed disk .................................................................................................................................... 53
Figure 2.31 Interference diagram for a 64 bladed disk ...................................................................... 54
Figure 2.32 Interference diagram for one family mode of a 40 bladed disk ...................................... 54
Figure 2.33 Mode coupling in interference diagram of a 12 bladed disk .......................................... 55
Figure 2.34 a) Example of shaft orbiting b) Tip leakage flow ........................................................... 57
Figure 3.1 Blade tip clearance and blade tip timing measurement system ........................................ 58
Figure 3.2 a) Typical sensor signal b) Influence of clearance variation ............................................ 59
Figure 3.3 The concept of the TOA variations .................................................................................. 61
Figure 3.4 Example of a) asynchronous and b) synchronous vibration ............................................. 63
Figure 3.5 a) Equally spaced sensors b) Non-equally spaced sensors c) Sampling of different
frequencies with non-equally spaced sensors .................................................................................... 63
Figure 4.1 Film of permalloy ............................................................................................................. 65
Figure 4.2 The Honeywell HMC 1501 .............................................................................................. 65
Figure 4.3 HMC1501 Signal output in function of theta ................................................................... 66
Figure 4.4 The sensor working principle ........................................................................................... 66
Figure 4.5 Left: Typical output value experimentally observed; Right: Sensor signal change with the
clearance [41] ..................................................................................................................................... 67
16
Figure 4.6 Test bench for the optimization of the sensor components layout ................................... 68
Figure 4.7 Top view of the bench and measurement grid .................................................................. 68
Figure 4.8 Peak to peak output signal values ..................................................................................... 69
Figure 4.9 The first magnetoresistive sensor prototype ..................................................................... 70
Figure 4.10 Typical modelling FEM mesh at different blade positions ............................................ 72
Figure 4.11 The magnetic induction change at the blade passage ..................................................... 72
Figure 4.12 Basic electronic circuit for static measurements ............................................................ 73
Figure 4.13 The test bench for the sensor characterization – version 1 ............................................. 74
Figure 4.14 The new disk and the little blade .................................................................................... 74
Figure 4.15 The bench equipped with a sensor shaking system ........................................................ 74
Figure 4.16 Left: the frame. Right: the assembly of the universal calibration bench ........................ 75
Figure 4.17 The Universal Calibration Bench for BTC & BTT systems .......................................... 76
Figure 4.18 The displacement monitoring sensor of the shaking system .......................................... 76
Figure 4.19 Typical signal collected with the sensor prorotype and a gear rotating in the lathe ....... 78
Figure 4.20 Detail of the calculated maximum peak values .............................................................. 78
Figure 4.21 Example of TOA calculation .......................................................................................... 79
Figure 4.22 Scheme of the TTOA and TOA calculation ................................................................... 80
Figure 4.23 Example of measured blade displacements .................................................................... 81
Figure 4.24 Example of waterfall of the measured blades displacements ......................................... 81
Figure 5.1 Left: scheme of the clearance test. Right: blade passage pulse at different clearances .... 84
Figure 5.2 Setup of the preliminary clearance test ............................................................................. 84
Figure 5.3 Acquired signals at different gaps and velocities ............................................................. 85
Figure 5.4 Detail of the peaks of tooth 3, 4 and 5 .............................................................................. 85
Figure 5.5 Maximum peak values variation....................................................................................... 86
Figure 5.6 The VKI R2 compressor rig ............................................................................................. 87
Figure 5.7 The calibration tower of the R2 compressor .................................................................... 88
17
Figure 5.8 The calibration curve of the magnetoresistive sensor for clearance measurements ......... 88
Figure 5.9 Detail of blade signals at different gaps ........................................................................... 90
Figure 6.1 The Campbell diagram of the test ..................................................................................... 92
Figure 6.2 The waterfall of the measured displacements of the first seven blades ............................ 93
Figure 6.3 Individual measured displacements of blade1 .................................................................. 93
Figure 6.4 3D Spectrogram of the individual measured displacements of blade1 ............................ 93
Figure 6.5 The Zplot of the measured displacements of blade 1 ....................................................... 94
Figure 6.6 The Fogale test bench ....................................................................................................... 95
Figure 6.7 A detail of the waterfall of the measured displacements with the capacitive sensors at
Fogale ................................................................................................................................................. 96
Figure 6.8 A detail of the waterfall of the measured displacements with the magnetoresistive
sensors at Fogale ................................................................................................................................ 96
Figure 6.9 The Zplot from the measured displacements with the capacitive sensors at Fogale ........ 97
Figure 6.10 The Zplot from the measured displacements with the magnetoresistive sensors at Fogale
............................................................................................................................................................ 97
Figure 6.11 (a-b) The SO-3 turbojet engine – ITWL, (c) The 4 slots in the casing to install the
sensors ................................................................................................................................................ 98
Figure 6.12 The waterfall of the measured displacements of the first 15 blades, ITWL sensors ...... 99
Figure 6.13 The waterfall of the measured displacements of the first 15 blades, magnetoresistive
sensors .............................................................................................................................................. 100
Figure 6.14 Comparison of the measured displacements around 12000 rpm .................................. 100
Figure 6.15 Comparison of the measured displacements around 12000 rpm, blade 1 .................... 100
Figure 6.16 Comparison of the measured displacements around 12000 rpm, blade 6 .................... 101
Figure 6.17 Left: the VKI R2 Campbell diagram. Right: the compressor inlet ............................... 102
Figure 6.18 The measured displacements in the R2 compressor, part 1 .......................................... 103
Figure 6.19 The measured displacements in the R2 compressor, part 2 .......................................... 103
Figure 7.1 The block diagram for the configuration of the streaming of data ................................. 105
18
Figure 7.2 The main block diagram ................................................................................................. 106
Figure 7.3 The front panel of the software ....................................................................................... 106
Figure 7.4 The differential magnetoresistive sensor ........................................................................ 107
Figure 7.5 The two signals from the differential sensor .................................................................. 108
Figure 7.6 Comparison of the calibration curves ............................................................................. 108
Figure 7.7 Raw measured displacements calculated with the matlab software ............................... 109
Figure 7.8 The Zplot of an aliased frequency .................................................................................. 109
Figure 7.9 The Zplot of the multiples of the aliased frequency ....................................................... 110
Figure 7.10 The Zplot after the anti-aliasing procedure .................................................................. 110
19
1 Introduction
1.1 Background
The engines currently installed in aircrafts contain limited instrumentation for monitoring the
engine condition in service. On line health monitoring sensors would allow improvements on safety
and, at the same time, reduce whole life costs and maintenance requirements. However, the aero
engines presents an extremely challenging environment for the application of reliable
instrumentation.
The accurate control of running gaps between static and rotating components is vital to preserve
mechanical integrity and ensure a correct functioning of any rotating machinery. In aero engines, tip
leakage above the airfoil tip also results in relevant aerodynamic losses. One way to measure and
monitor tip gap is by a Blade Tip Clearance (BTC) technique. The clearance variation occurs as a
consequence of manufacturing and installation tolerances, vibrations, different expansion rates of
blades, rotor disk and casing induced by thermal and speed transient [1].
The measurements of vibration and eigen frequencies is a crucial part of the development of any
new gas turbine or aero-engine. All turbomachinery blades and components experience vibration
during operation due to several sources like: unbalance, pressure distorsions, flutter, stall and surge.
Unfortunately this undesirable structural motion affects fatigue life, performance and integrity [2].
For more than half a century, vibration measurement has been performed by installing strain gauges
on the blades and using telemetry to transmit the signals. The Blade Tip Timing (BTT) technique, is
currently being adopted by all manufacturers as a replacement for the classical strain gauge
technique because of its non-intrusive character. The aim of the blade tip-timing measurement
system is to evaluate the time at which a point on a rotating blade tip passes a stationary point. In
the absence of any structural vibration, the blade arrival time would depend only on the rotational
speed of the blade. On the other hand, when the tip of the blade is vibrating, the arrival time will
depend on the amplitude, frequency and phase of vibration. Hence, the blade motion can, in
theory, be determined from the arrival data.
Various types of sensors have been proposed for BTC and BTT systems including electromagnetic,
magnetic, inductive, optical, capacitance, eddy currents and microwave [3]. The engines present an
extremely challenging environment for the application of these monitoring sensors but still now, no
one reliable and accurate in service system exist.
20
There are several requirements that such an instrumentation should satisfy to be used in a modern
engine: small sensor size, freedom from geometry sensitivity, measure of the real clearance (casing
- blade tip), tolerance to debris, suitable for high temperature measurements, reliable, long life [4].
About 10 years ago, Von Flotow [5] reports a list of patents filed by various developers of blade tip
sensing systems, this demonstrates the level of interest and activities in that field.
The aim of this project was to develop and test a new sensor for clearance and vibration
measurements in turbomachines, based on magnetoresistive sensing elements. This research work
started from the consolidate experience and the state of the art on non-intrusive measurement
systems developed by the research groups of the Universities of Perugia and Ancona, in
collaboration with important turbomachine manufacturers like the ex-Nuovo Pignone. In fact, Prof.
E.P. Tomasini, prof G. Rossi, prof. N. Paone and the Engineer P. Nava have been working on this
topic since the first 90s’ [65-66]. Beside that, couple of years before this research project has
started, the University of Perugia was already testing the digital magnetoresistive sensors [67].
1.2 Literature review
The literature survey has started from the VKI LS on Tip Timing and Tip Clearance problems in
turbomachines [6]. Then, some information on vibration and clearance effects were gathered from
the library. Starting from the VKI LS several references about BTT and BTC application systems
were found, mainly collecting the most important, interesting and some recent publications with
different sensor technologies. Improvements and new publications from authors were also collected.
The next step was performed searching for BTT data analysis methods. It was not easy to find
information on this topic because of confidentiality issues. Finally the last part of the bibliographic
research was focused on the magnetoresistive sensing elements and related applications, with
particular attention to the aero engine.
1.2.1 BTC and BTT measurement systems
Various kind of sensors have been proposed to measure the gap between blade tip and casing as
well as blade vibrations. Chana in [3] gives an overview of the main type of sensors: optical,
capacitance, eddy current, microwave and pressure. Optical probes are largely used for blade tip
timing thanks to the fast response achievable and very small spot size. Cardwell [7] used an optical
system to compare the result with an eddy current sensor. Gallego [8] and Heath [9] used optical
systems to validate a new timing technique. Moreover, optical probes were also used for blade tip
21
clearance as reported by Dhadwal et. al [10] and Pfister [11]. Zielinski [12] reports a capacitance
based tip timing system for crack detection on an industrial gas turbine. He made a comparison
between the result of capacitance sensors and reliable data from strain gauges. A recent work at
VKI with the same type of sensor was made by Lavagnoli et al [13]. They reported that blade tip-
casing gap can be measured within 25 μm (20:1) at a working distance of 1 mm. General
Dynamics Advanced Technology Systems [14] have patented a three coil differential and shielded
eddy current sensor for tip timing. Chana [7] has recently developed and tested eddy current sensors
on an aero engine. Tests have been performed using sensors mounted on Rolls-Royce and QuinetiQ
facilities. Grzybowski [15] and Maslovskiy [16] presented a system based on microwave sensors for
clearance measurements. The same type of probe was also used by Wagner et. al [17] for vibration
measurements. Detailed measurements of the pressure field on the casing of a turbo-machine can
also potentially provide information about the tip clearance and tip timing. Belsterling [18] in 1971
suggested using a non-contact pressure measurement method for determining blade tip gap and
displacement.
1.2.2 Measurements without holes
Measurements done through the casing wall without making holes in it would be very
advantageous. Some interesting articles related to this topic have been found. Rickman [19]
reported in 1982 the use of a commercially available eddy current sensor to measure blade speed
through the casing of a diesel engine's turbocharger for use as a fault detector. More recently, Von
Flotow et al. [20] tested the response of an eddy current sensor through different shielding materials
in the thickness range below 2mm. They reported that titanium and stainless steel did not affect the
sensor significantly while aluminum reduces the signal by approximately half. Finally, also Chana
[7] confirmed the possibility to measure through different shielding materials with the same kind of
sensor, obtaining an acceptable signal attenuation for a few mm thickness.
1.2.3 Blade Tip Timing Data Processing
Tip timing measurements consist of three steps as follows:
1. Data acquisition: the acquisition of blade time of arrival data from a number of probes
mounted around the casing
2. Data postprocessing: the derivation of characteristic vibration parameters such as blade
displacement from the measured data
22
3. Data analysis: the analysis of the sampled blade tip deflection to retrieve the vibration
properties of the rotating assembly
Data postprocessing consists first in identifying surplus or missing blade readings in the data file
and to correct it. Then to process the timing information into blade displacements. Details are given
by Gallego [8, 21] and Carrington [22]. Finally, the step 3 is crucial as it provides important
information about the dynamic behavior of the assembly. In this project, a methodology based on
the cross-correlation function has been developed to process the tip-timing data. Further detail will
be given in this report and can also be found in previous work by Prof. Gianluca Rossi, during his
PhD Thesis [23]. A number of methods have been proposed in literature like the single parameter
technique developed by Zablotskiy and Korostelev [24], the two-parameter plot method (Heath et al
[25, 26]), the technique by Zielinski and Ziller [27] ,and various autoregressive methods
(Dimitriadis et al. [28], Carrington et al. [29], Gallego [30], Gallego and Dimitriadis [31]). The
processing is further complicated when two or more modes of vibration exist simultaneously. This
is not an uncommon event and has been repeatedly reported in the literature regarding aeroengine
vibration problems since the early investigations of Armstrong and Stevenson [32]. Autoregressive-
based methods and the two-parameter plot method have been shown to fail when two simultaneous
synchronous resonances from simulated BTT data are analysed (Gallego [30]). Robinson [33, 34]
reported the existence at Pratt and Whitney of techniques for decoupling two simultaneously
occurring synchronous modes. This vibration condition is also referred to (Jones [35] , Hayes et al.
[36]). However, no description of the methodology is given by the mentioned authors. Also Gallego
presents some methods for the analysis of BTT data from assemblies undergoing two simultaneous
resonances [37]. Having a rotor blade vibrating at two modes and considering the displacement of
the tip as a linear superposition of two undamped sinusoids, it is possible to determine the
autoregressive coefficients. Than, the frequencies of vibration can be recovered. The linear
autoregressive model requests the rotational speed of the rotor to be constant or varying very slowly
in the period of the computation. This method can be used both for data acquired in one revolution
or in multiple revolutions. In the latter case the method has been called MGARES. After an
intensive experimental validation of the proposed methods, Gallego [8] recommended MGARES as
the method for the analysis of BTT data from assemblies undergoing single and multiple
simultaneous resonances.
23
1.3 Objective of the research project
The specific objectives of this research work derive from the intention of designing and realizing a
BTC and BTT measurement system based on new magnetoresistive sensing elements. These
objectives can be summarized as:
1. design and realization of a new probe;
2. development of dedicated hardware and software components;
3. characterization and experimental investigations of the whole instrumentation chain ;
4. comparisons and validations with existing systems.
1.4 Structure of the thesis
The work reported in this thesis has been divided into eight sections.
Chapter 2 introduces the theory of vibration in rotating components. Particular attention is given to
blade and disk vibration modes. Finally, the phenomenon of blade tip clearance variation is
described.
Chapter 3 explains the principle on which a Tip-Timing and Tip Clearance system works.
The design, development and realization phases of the measurement system are reported in chapter
4.
Chapter 5 and 6 focus on experimental measurements of tip-clearance and tip-timing. The
characterization of the sensor for clearance measurements and its calibration curve is illustrated.
Blade vibrations were simulated and measured on test benches and real turbomachines.
Some aspects of the whole measurement chain were improved during this work. Chapter 7
summarizes the main developments.
Finally, conclusions from this research project are drawn in Chapter 8. The main features and
performances are summarized. The thesis ends with recommendations for future work.
24
2 Fundamentals of blade vibration and blade tip clearance problems in
turbomachinery
2.1 Introduction
High cycle fatigue of blades due to dynamic loading during vibration is the main cause of engine
failures. Understanding the mechanism of these vibrations forms an essential basis for a designer to
avoid a design operating with dangerous vibrations. This chapter goes stepwise through the possible
vibrations of a bladed disk. First the blade and disk vibrations are explained separately, finally the
combination of both vibrations is handled. The chapter ends with an overview on blade tip gap
variation problems.
2.2 Blade modes
2.2.1 Classification of blade modes
Vibrations take place in the neighbourhood of certain frequencies, called eigenfrequencies, which
are typical for the vibrating object. Each eigenfrequency has a characteristic mode shape (=
eigenmode). This is the shape of the vibrational pattern. Mode shapes of a blade are typically
divided into four families, [38]: flap wise bending (F), edgewise bending (E), torsional bending (T)
and an axial mode (A).
Since blades usually have a very complex shape, it is common to investigate their vibratory
properties using three simplified models according to their geometric characteristics such as length,
thickness and chord, [38]:
(a) If the length-to-chord aspect ratio (=Length/Chord) is high (larger than 5) and the thickness is
small compared to the length, then the blade is considered as a uni-dimensional structure and
modelled as a beam.
(b) If the length-to-chord aspect ratio is low and the thickness is still small compared to the other
dimensions, then the blade is considered as a two dimensional structure and modelled as a thin plate
or shell.
(c) If the blade has a low length-to-chord aspect ratio and is considered as being thick, then it is a
three-dimensional structure. In this case, the blade is modelled as a thick shell.
25
Figure 2.1 First 3 bending modes of beam with one end fixed and one end free
Figure 2.2 First 3 bending modes of beam with both ends fixed
The simple beam model (a) has the simplest vibration characteristics. Let’s consider two beams
with different end conditions that can correspond to the situation of a realistic blade. The first beam
has a clamped and a free end, which corresponds to an unshrouded blade. The other beam is
clamped at both ends and represents a shrouded blade. Figures 2.1 and 2.2 represent their first three
bending modes. Notice the points with no deflection, they are called ―nodal points‖.
The mode shapes of thin and thick shells are more complex, because the motions can take place in
three dimensions. Figure 2.3 depicts the mode shapes of the first flap wise, edgewise and torsional
bending mode. Even more complex mode shapes are formed when two or more modes appear
26
simultaneously. It is clear that determining the type of mode shape (especially complex ones), is not
an easy task.
Figure 2.3 Mode shapes of cantilevered thin shells a) First flap wise bending mode; b) First torsional mode ; c) First edgewise
bending mode
Figure 2.4 Nodal lines of modes shapes from cantilevered blades
Specifying nodal lines is generally found as the most convenient way to describe mode shapes.
These lines represent points which have no displacement during vibration, see figure 2.4. When
investigating the vibratory behaviour of a blade, it is always useful to look first at the nodal lines.
They give quick information about the mode shape and the places where strong deflections are
expected. Figure 2.5 is an example of mode shape classification from a cantilevered blade by means
of the nodal lines, [39]. Two types of nodal lines are indicated. The horizontal lines represent m
radial directed nodal lines and the vertical lines are n axially or tangentially directed nodal lines.
27
Note that in all shown cases, a vertical nodal line lies on the line where the blade is fixed to the hub.
The blades in the first and the third column (m=0 and m=2) are vibrating in bending modes. The
middle column (m=1) has blades that are subjected to a torsional mode.
Figure 2.5 Mode shape classification of a cantilevered blade
Figure 2.6 Non-vibrating radial compressor blade
28
Figure 2.7 Two vibrating modes of a radial compressor blade
Figure 2.8 Two vibrating modes of a radial compressor blade with nodal line indication
The strength of this method can be illustrated by analysing the complex mode shapes of a non-
rotating radial compressor blade. The hub is simulated by applying fixed boundary conditions on
the blade’s root. Finally, the mode shapes are obtained with a Finite Element Program. Figure 2.6
illustrates the non-vibrating blade, whereas two mode shapes (without nodal lines) are given by
figure 2.7a and b. In figure 2.8, nodal lines are formed by the intersections of the blade at rest and at
a vibrating condition. The nodal lines clearly indicate the (m,n) parameters of the modes. Figure
2.8a with (m,n) = (1,2) is a torsional mode. Figure 2.8b with (m,n) = (3,2) is a bending mode.The
blades in figures 2.6, 2.7 and 2.8 are modelled as thick shells. A step closer towards real blades
would be a model that uses tetrahedral blocks. This 3D model type seriously increases the
computational time for finding eigenfrequencies and modes, but reaches a higher accuracy. In most
cases (simple blade shapes) the gain in accuracy is not worth the larger computational time
29
compared to the previous presented models. On the other hand, since blade failures are mostly
caused by metal fatigue due to vibrations, it remains very important to predict them as accurate as
possible.
2.2.2 Parameters affecting the blade mode properties
The effect of the blade shape:
- Taper affects the values of the eigenfrequencies but doesn’t change anything to the mode shapes.
There is though a difference between tapering the chord or the thickness, see figure 2.9. A decrease
in chord gives an increase in both bending and torsional mode frequencies. Giving taper to the
thickness (decreasing the thickness) results into, [39]:
• An increase of the first bending frequency
• A decrease in frequency for all other higher bending modes
• A decrease in frequency for all torsional modes
Figure 2.9 Taper of blades in a) thickness and b) chord
- Adding twist appears to have the opposite influence, namely same eigenfrequencies and different
mode shapes for the bending modes. The latter is caused by combinations of blade modes that
interact. Giving more complexity to a blade increases the chance that different mode shapes will
influence each other. As the pre-twist increases, the blade’s torsional stiffness increases. Therefore,
torsional frequencies will increase with increasing pre-twist angle, [39].
30
- All blade vibration frequencies appear to increase almost linearly with increasing blade thickness,
[40].
- A centrifugal compressor blade outlet angle has great influence on the mechanical and dynamic
characteristic of a blade. When decreasing the outlet angle from 60 down to 0° (backswept blades,
where 0° is a radial directed blade), the stress due to centrifugal force in the blade outlet part at the
hub diminishes with 40%, see figure 2.10, [41].
Figure 2.10 Equivalent stress in blade outlet part at hub due to centrifugal forces
Figure 2.11 First bending frequency of a blade as function of the blade impeller
- The influence of the blade outlet angle on the dynamic behaviour of a centrifugal compressor
blade is shown in figure 2.11. Stiffening of the structure for the first bending mode is indicated by
31
the change in eigenfrequency. The example in figure 2.11 achieves a maximum stiffness at
approximately 30°, [41].
The effect of operating conditions, [38]:
- Rotating blades are submitted to a centrifugal force. Large twisted blades can suffer from a change
in their geometry due to this force. However, the most important influence of the centrifugal force is
a stiffening of the blades. This causes in turn an increase of the natural frequencies with some
corresponding changes in the mode shapes. In figure 2.12 the basic feature of centrifugal stiffening
is shown by a centrifugal force Fc pulling at the non-vibrating blades. When a blade starts to
vibrate, it’s gravitational point will move due to the blade’s motion. This shift causes Fc to create a
moment around the blade’s root. One can see that the centrifugal force stabilizes vibrations normal
to the rotation plane (2.12 b), because the moment counteracts the motions. Vibrations in the
rotation plane (2.12 a) have an additional destabilizing moment. This is due to the changing
direction of Fc, during the blade’s motion, which creates a small force component in the same
direction as the motion of the vibrating blade.
- Next to the stiffening effect due to the centrifugal force, is the de-stiffening effect (= a lowering
of the eigenfrequencies) due to a temperature increase. Since the temperature, of a fluid passing a
blade, usually increases with increasing operating speed, it is possible that both effects will
compensate each other. This is certainly the case for high pressure compressors and low pressure
turbines. In case of a high pressure turbine and a low pressure compressor, respectively the thermal
and rotation effect will dominate. This behaviour is strongly dependent on the blade material
properties.
32
Figure 2.12 Centrifugal stiffening: a) tangential vibration; b) axial vibration
2.3 Disk modes
A disk has, similar to blades, eigenmodes and eigenfrequencies. The vibration of a disk is
characterised by circumferential and radial sinusoidal displacements on respectively concentric
circles and radial lines. The nodal points of these displacements form respectively nodal diameters
(ND) and nodal circles (NC). The disk motions for modes ranging from 0 up to 3 ND (with 1 NC on
the circumference) are depicted in figure 2.13 a. Figure 2.13b shows the displacements for 1 up to 4
NC modes without nodal diameters. Combinations of nodal diameters and nodal circles result in
complex shapes as depicted in figure 2.14, [42]. These sinusoidal patterns are possible in axial,
radial and tangential direction, see figure 2.15, [43].
33
Generally, when the frequency goes up, the number of nodal diameters (and circles) increases.
Nodal diameter modes are grouped into families, where each family is characterized by a mode
shape (axial, tangential or radial) with one specific number of nodal circles. Figure 2.16 shows the
eigenfrequency behaviour from three families (0 NC, 1 NC and 2 NC) having the same mode shape.
It is common to present the response of diametral modes (no NC) on concentric circles. The disk
rim, which has the largest displacements, is most suited for this purpose, see figure 2.17.
Figure 2.13 Deformation of disk with a) only ND(1 NC) and b) only NC
34
Figure 2.14 Axial deformations of a disk with m = number of ND and n = number of NC
Figure 2.15 Disk vibrations
35
Figure 2.16 Families of a disk mode
Figure 2.17 Presentation of 4 ND mode with displacements in axial direction
36
2.4 Bladed disk modes
Figure 2.18 First bending modes of blades coupled through a disk: a) in-phase vibration; b) counter current vibration; c) one
blade at rest
When combining the blades with a disk into a bladed disk, vibrations of both components must be
taken into consideration, [44]. Because all blades are connected to each other through the disk and
optionally a shroud, a coupling between the blades will exist. This blade coupling creates bladed
disk modes which are characterized by the motions of the blades relative to each other. At a given
moment, all blades will have the same mode shape but with different amplitude, see figure 2.18.
Nodal diameters and circles are also used to categorize bladed disk modes. For instance, the first
tangential blade bending mode (= 1T) is possible on a bladed disk for several numbers of nodal
diameters. One can group these blade coupling modes together into one family. On their turn, each
family has a specific number of nodal circles. This means that many families are possible for, in this
case the first tangential blade bending mode. For instance, the 3 different modes, although the
37
blades have the same mode shape, of the bladed disk in figure 2.18 are part of one family if they all
occur with the same number of nodal circles. Considering the modes within one family: With 0
nodal diameters all blades are moving in phase, but starting from 1 nodal diameter the blade
amplitudes follow a sinusoidal pattern. The points of no displacement (where the sinus is zero) are
on a nodal diameter. When a nodal diameter lies on a blade, then the amplitude of its blade mode is
zero (the entire blade is at rest). If a nodal diameter would lie in between two blades, then both
blades will have an opposite sign for the amplitude of their mode shape. For a bladed disk with an
even number of N blades there can be maximum N/2 nodal diameters, if N is odd the maximum is
(N-1)/2. Higher numbers of nodal diameters are possible but they are indistinguishable. An example
of a 36 bladed disk, where the 4th
and 32nd
diametral patterns are indistinguishable, is given in
figure 2.19. Note that the dots in this figure indicate the position of the blades at a certain moment,
thus at every moment the position of the blades for a 4 ND pattern is the same as for a 32 ND
pattern. Although the 32 ND sinusoidal pattern has a higher frequency then the 4 ND pattern, the
blades experience a vibration as would it be a 4 ND pattern at an eigenfrequency of the 4 ND
pattern.
Figure 2.19 Indistinguishable 4th and 32nd nodal diameter pattern of a 36 bladed disk
38
Figure 2.20 Disk stiffening effect caused by increasing number of nodal diameters
Figure 2.21 Effect of disk stiffeness on the bladed disk modes
39
Generally, the number of nodal diameters increases with increasing frequency within one family.
The only exception is the 0 nodal diameter mode which is in most cases harder to achieve than the 1
and 2 nodal diameter patterns and therefore has a higher frequency. Important to notice, as the
number of nodal diameters increases (up to N/2 for N even), the natural frequency in each family
approaches one of the cantilevered blade frequencies, see figure 2.20. The latter is the natural
frequency of a single blade with one root fixed in case of an unshrouded disk. This behaviour is
caused by the progressive stiffening of the disk as it adopts more complex shapes (= more nodal
diameters). In case of a very stiff disk, the coupling between the blades is very low, so the natural
frequencies are very close, even at small numbers of nodal diameters, to the blade alone
frequencies. The stiffness effect is illustrated in figure 2.21. An increase in number of nodal circles
will also make the disk stiffer. This stiffening has the same influence as the increase in nodal
diameters, but it will additionally cause these families (starting from the first nodal circle) to occur
at higher frequencies. There is a very important difference in handling a bladed disk, where blades
and disk have or don’t have comparable stiffness. In case they do, there will be a strong blade
coupling and a dynamic analysis of the whole disk is necessary to provide accurate estimates of the
natural frequencies and mode shapes. However, when a disk is very stiff compared to his blades,
it’s influence can be disregarded and a dynamic analysis of a single blade is sufficient, because only
the blades are vibrating.
Figure 2.22 Nodal diameters in bladed disk with 6 blades in each disk sector
40
Figure 2.23 Mode shapes of one disk sector where a black arrow denotes the position of a nodal diameter
When blades are grouped into sectors, for example six blades in one sector, the vibrational character
of the bladed disk is even more difficult to evaluate, [45]. Figure 2.22 illustrates bladed disks, where
each blade belongs to a sector of 6 blades. As long as the nodal diameters lie in between adjacent
sectors, all blades have the same displacements within each sector, see figure 2.22. When a nodal
diameter intersects a sector, new modes are created within this sector. Different modes for an
individual sector are shown in figure 2.23. All blades are vibrating with the same blade mode, but as
one or more nodal diameters (indicated with black triangle) intersect a sector, the amplitudes of the
blade modes are changed.
Single and double modes:
Each circumferential symmetric bladed assembly has certain well defined types of vibration modes.
The most important feature is the existence of two mode types in one family, namely single and
double modes. The latter one is the most common. A double mode consists actually of two
orthogonal harmonic modes (a sin and a cos) with the same natural frequency. There is no unique
vibration pattern that can be specified for these double modes. One should consider two orthogonal
41
shapes and when the structure is vibrating at that natural frequency, the structure can assume any
form given by a linear combination of the two specified shapes. Suitable shapes are given by
cos(nθ) and sin(nθ). These are circumferential distributions of displacement around the assembly
with n nodal diameters. So at the natural frequency, the assembly can vibrate in any combination of
these two patterns, that is in a shape of the form cos(nθ + ϕ). This implies that the assembly vibrates
with n nodal diameters whatever the bladed disk mode shape may be. The orientation of these
diameters depends on some additional external influence. It is thus possible that the nodal diameters
lie on a blade or somewhere in between two blades, [44].
The single modes, only one mode at resonance, correspond to motions with either all the blades in
phase (0 nodal diameters) or out of phase (N/2 nodal diameters, only possible if N is even).
Important remarks:
The 0 and 1 nodal diameter modes at the lower end of each family have a special characteristic.
These three modes (one single and one double) differ from all the other modes (two or more nodal
diameters) because they have a net motion of the centre of the disk. A 0 nodal diameter mode
involves an axial or a torsional disk movement (see movement of the blades in figure 2.24c) while a
1 nodal diameter mode has translation of the disk along a diameter (see figure 2.24b). Therefore
these three modes are also influenced by the shaft and bearings which support the disk. One should
include these components to have more accurate estimates of the 0 and 1 nodal diameter modes,
[44]. One must also take into account a second coupling between the different mode shapes. A
coupling is likely to happen when both the frequency and the number of nodal diameters of two
different mode shapes (for example first bending and first torsion) are close to each other. The
influence of coupling is quite weak for modes that only represent pure bending or torsion, but it is
significant for highly coupled modes like the FE-T (flapwise + edgewise + torsional) mode, [38].
Figure 2.24 Blade motions during vibration: a) No vibration, b) 1ND with translation along diameter, c) 0ND with torsional
disk movement
42
2.5 Bladed disk response
2.5.1 Engine order excitations
Excitations of a bladed disk are very closely related to engine orders (EO). The latter are harmonics
(natural multiples) of the rotation speed. For example, a rotation speed of 3000 RPM is equivalent
to (3000/60) rates per second = 50 Hz. Thus the first harmonic (=1 EO) is 50 Hz, the second (=2
EO) 100 Hz, the third (=3 EO) 150 Hz, etc. One pressure fluctuation (caused by for example a
vane) during one rotation will be experienced by an observer rotating with the disk, as a pulsation of
first EO. For instance, 5 vanes will cause pulsations of the 5th
EO, because the observer feels 5
pulsations per rotation. Sometimes these pulsations have strong higher Fourier components, thus
previous example may also experience pulsations of the 10th
, 15th
, 20th
…EO. It is proven from tests
that these Fourier components weaken as the component number increases.
2.5.2 Response of a bladed disk
Two angular coordinate systems are defined to describe the response of a bladed disk, [46]. The
first one is a stationary reference frame where θ0 represents the angle. The other reference frame is
fixed to the disk rotating with rotational speed Ω (rad/s) and uses the angle θ. The relation between
both coordinates is given by: θ0 = θ − Ω t since the direction of the rotation and the angles are
chosen opposite to each other.
Excitations with a certain EO are mostly initiated by pressure distortion fields of, for instance, S
distortions located in the flow field. Generally, these pressure distortion fields are equally spaced on
the circumference and do not have abrupt changes. Therefore, they can be represented as a
sinusoidal forcing function V with amplitude A/2 (see figure 2.25), [47]:
( )
( ( )) (2.1)
The bladed rotor has M blades which transfer the forcing function V to the disk. It is thus very
likely that they affect the nodal diametral patterns. The presence of the blades is approximated by a
Fourier series in circumferential direction. For simplicity only the first component of this series is
used such that blades are represented by:
( )
( ( )) (2.2)
43
Figure 2.25 Sinusoidal distortioned flow due to the stator vanes (upper graph) and the sinusoidal representation of the rotor
blades (lower graph)
The actual force F, felt by the blades, is then:
( ) ( ) ( ) (2.3)
Or:
( )
o ( ( ))
o ( ( )) o ( ) (2.4)
The double mode shapes of the nth
diametral mode are assumed to be ϕ1(θ) = cos(nθ) and ϕ2(θ) =
sin(nθ) with eigenfrequencies ωn1 and ωn2. Studying the response of a bladed disk in it’s nth
diametral mode due to a force F is commonly done by using generalized forces N1 and N2 (for
respectively the cosine and sine mode). The generalized force for each assumed mode is calculated
with:
∫ ( ) ( )
(2.5)
Or:
∫ ( ) ( )
(2.6)
44
∫ ( ) ( )
(2.7)
Substitution of the force F from (1.4) into (1.6) gives:
o ( )) ∫ o ( ) ( )
o ( ) ∫ o (( ) ) ( )
o ( ) ∫ o (( ) ) ( )
(2.8)
The terms in this expression are always zero except for the following numbers of nodal diameters n:
n = S
o ( ) (2.9a)
n = M - S
o ( ) (2.9b)
n = M + S
o ( ) (2.9a)
Similar for the generalized force N2:
n = S
in( ) (2.10a)
n = M - S
in( ) (2.10b)
n = M + S
in( ) (2.10a)
If a very lightly-damped system is assumed the generalized responses are calculated with a
simplified convolution integral:
( )
∫ ( ) ( ( ))
(2.11)
( )
∫ ( ) ( ( ))
(2.12)
Where m is the modal mass or generalized mass of the mode. After integration, the generalized
responses are found to be:
( ) o ( ) o ( ) (2.13)
( ) in( ) in( ) (2.14)
45
The parameters Wii depend on the values of the generalized forces N1 and N2. On their turn, they
depend on the number of nodal diameters (1.9 and 1.10). Therefore, in case the number of nodal
diameters is equal to the number of stator vanes (n = S), the parameters Wii become:
( )
( )
( )
( )
Note that all Wii can also be found for n = M − S and n = M + S.
One can see from the equations of the generalized responses (2.13 and 2.14) that once the system is
set into motion, it will tend to vibrate at its natural frequency as well as to follow the frequency of
the excitation. Since the system is (lightly) damped, the part of the motion not sustained by the
sinusoidal excitation will eventually die out. This is the transient motion which describes the
oscillations under free vibration at the natural frequency of the system. The motion sustained by the
sinusoidal excitation is called the steady-state response.
Hence the system vibrates at the excitation frequency. It is thus possible to drop the second terms in
the expressions of the generalized responses, because they represent a transient response. Therefore,
the steady-state solution of the responses is:
( ) o ( ) (2.15)
( ) in( ) (2.16)
The total response is calculated by the mode summation formula:
( ) ∑ ( ) ( ) (2.17)
Or:
( ) o ( ) o ( ) in( ) in( ) (2.18)
Here the result is considered for a perfectly tuned bladed disk, thus when ωn1 = ωn2 and mn1 = mn2
the response can be written as:
( ) o ( ) (2.19)
46
Remember that W11 only exists (≠0) when n = S, n = M − S or n = M + S. Since the disk becomes
stiffer with increasing nodal diameters, it is unlikely that strong excitations will occur for the case n
= M + S. Only investigation of the other two cases is relevant:
-Case: n = S then:
( ) o ( ) o ( ) (2.20)
-Case: n = M - S then:
( ) o (( ) ) o ( ) (2.21)
In the first case, the n diametral pattern always travels with the same speed but in the opposite
direction of the disk rotation. The response can be seen as a standing wave in the stationary
reference frame.
For n = M − S, the disk response has a diametral pattern that can travel in opposite or same
direction as the rotation, depending on the values of M and S (whether n is < 0 or > 0). Be aware
that a rotating reference frame is used, so the diametral patterns will only rotate slower or faster
than the disk speed when they are seen in a fixed reference frame.
Note that there is also danger for 2M − S, M − 2S, 2M − 2S, 3M − S,… modes which are found if
higher Fourier components were used for the mathematical representations of the blades and the
distortions.
2.5.3 Classification of bladed disk response
In general a distinction is made between two response modes X(θ, t) when resonance occurs, [48].
They differ according to whether the number of rotating blades has an influence or not. In both
cases, the initiation is due to a number of stationary elements S.
Critical disk speed resonance
The response for this type is similar as in formula 2.20. The conditions which must be
simultaneously fulfilled are written below:
{
( ) ( ) ( )
(2.22)
47
With: y integer value, typically 1 and 2, higher values give much lower response
M number of rotating blades
S number of stationary elements ex. vanes
h harmonic of speed = EO
fr natural eigenfrequency during rotation rotational speed in Hertz
n number of nodal diameters
These conditions demand a bladed disk with at least two blades to transfer the pressure fluctuations
(forces) to the disk (a). The eigenfrequency of the mode must be equal to the frequency of the
excitation (c). For the last but most important condition (b), the bladed disk motions must be in
phase with the effective sinusoidal pattern of the excitation source. When the latter condition is not
fulfilled, then the effective forces would cancel out in a revolution. This is illustrated in figure 2.26,
where an n diameter mode shall be resonant with n and not with n+1 sinusoidal pulsations per
revolution.
Example:
It is disconcerning for a 4 nodal diameter mode ( n = 4 ) of a bladed disk to place 4 stator vanes ( S
= 4 ) in the inlet. At a certain rotational speed ( = disk critical speed ), condition (c) will be fulfilled.
Since the vanes produce 4 pulsations ( h = 4 ) during one revolution, the first Fourier component ( y
= 1 ) of these pulsations fullfills condition (b): 1S = h = n. The same can be said for the second
Fourier component ( y = 2 ) of 2 vanes in the inlet, because then 2S = h = n. Note that in this
example both critical disk speeds are equal.
Blade/vane interaction resonance
This type is defined as resonance due to the presence of blades. Since blades transfer pressure
fluctuations to the disk, there will exist some resonance at certain number of blades. For this type of
resonance, the following three conditions must be simultaneously fulfilled:
{
( ) | | ( ) ( )
(2.23)
48
with z an integer value, typically 1 and 2, higher values give much lower response. From condition
(a) one can conclude not to use equal or multiple numbers of blades and vanes. In that case no circle
modes (NC) can be excited in the 0 (= n) nodal diameter mode, because the 0 ND mode cannot
exist ( yS ≠ zM ). Remark that condition (b) is there to emphasise that resonance is at a harmonic
related to the number of stationary elements and not to the number of rotating blades.
Example:
Consider the example of a bladed disk with 27 blades and 16 stationary vanes. For the first
resonance type, two critical disk speeds should be investigated at the 16th
and 32nd
EO for
respectively the same nodal diameter modes. These EO should also be investigated for the second
type of resonance. The |2S−M| = |2x16−27| = 5 and |S−M| = |16−27| = 9 nodal diameter modes will
interact (at certain rotational speed) with respectively the second harmonic (h = 32 EO) and the first
harmonic of the vanes (h = 16 EO).
Figure 2.26 Nodal diameters and excitation of the same and not the same EO
49
2.5.4 Single and double modes
All diametral modes of a bladed disk consist of single and double modes, [49]. The response of a
bladed disk at a disk critical speed Ω (rad/s), where the disk resonates in an n diametral mode to an
nth
EO excitation (n = S), was found to be (formula 2.18):
( ) o ( ) o ( ) in( ) in( ) (2.24)
Assuming a perfectly tuned system:
X0 = W11 = W21 (2.25)
Thus:
( ) o ( ) o ( ) in( ) in( ) (2.26)
One can consider each of these two terms being the combination of a harmonic excitation at the
same frequency and phase of all blades (sin(nΩt)), but with the amplitude depending on the position
on the disk in a sinusoidal pattern (X0 sin(nθ)). Both terms of X(θ, t) represent a standing wave in a
rotating reference frame. These standing waves are the double modes of a diametral response. An
example of standing waves with 2 nodal diameters from a 12 bladed disk can be seen in figure 2.27.
Combining the standing waves results into a travelling wave (or a standing wave in a non-rotating
reference frame):
( ) o ( ( )) (2.27)
Figure 2.27 Nodal diameters of orthogonal standing waves of a 12 bladed disk
50
Considering now the zero and N/2 nodal diameter modes of a tuned bladed disk. The zero diametral
mode is found from (2.19) with n = M − S = 0 and X0 = W11:
( ) o ( ) o ( ) (2.28)
There is no phase difference between the responses of the blades and the total response is one
standing wave. Thus all blades have the same deflection at each time.
An excitation of an N/2 diametral mode means that the phase difference between adjacent blades is
180°. Suppose that the blades coincide with the nodal points of sin (nθ), then each nodal diameter of
cos (nθ) lies in between two adjacent blades. The second term in (2.18) is always zero, such that the
response for each blade with blade number n (n = 1 → N) can be written as :
( ) o ( ) (2.29)
This expression also represents one standing wave.
Both, the zero and N/2 nodal diameter mode, consisting out of one standing wave (in a rotating
reference frame) are classified as single modes.
2.6 Campbell diagram
Campbell diagrams are used to find rotational speeds which can lead to dangerous excitations of a
bladed disk, [50]. Since a diagram of the latter is quite complicated, it is more appropriate to
investigate a diagram of an isolated single rotating blade (no blade coupling possible), see figure
2.28. This diagram has the rotating speed (rpm) as x-coordinate and the frequency on his y-axis.
The lines with (1F, 2F, 1T, 3F) represent the eigenfrequencies of the mentioned blade mode shape.
Note how the stiffening of the centrifugal force affects the eigenfrequencies. The de-stiffening
effect of the temperature is not included. A second type of lines are the EO (=Engine Order) lines.
These lines are straight, because 1 EO = 1 pulsation per rotation times the RPM/60, 2 EO = 2
pulsations per rotation times the RPM/60, etc.
51
Figure 2.28 Campbell diagram of blade on rotating disk
Intersections of both line types are dangerous operating points, since they stand for equal
frequencies of an excitation source and a mode shape. In a reduced operating range starting at 4000
RPM up to 5500 RPM, all (3) intersections are highlighted. Note that, for example, the critical
speed of a 2nd
bending/3EO intersection requires 3 distortions within one rotation. This Campbell
diagram warns to certainly avoid having 3, 5 or 7 distortions during one rotation. For instance, four
vanes in front of this rotating disc would give no problems for the blade in this operating range,
since there is no intersection of the 4 EO line with an eigenfrequency line.
Everything explained up to now is valid for a single blade, but problems occur when extending this
diagram to a bladed disk. For a bladed disk, in addition to the frequency, also the mode shape of the
excitation source is very important for resonance. Therefore, the nodal diameters are needed. In
figure 2.29 the Campbell diagram is drawn from a 36 (= N) bladed disk. All horizontal lines in this
figure represent the eigenfrequencies of one and the same bladed disk mode but each line has a
52
specific number of nodal diameters going from 0 ND (lowest line) up to N/2 ND (upper line). Thus
at every horizontal line the blades are vibrating in, for example, the first flapwise bending mode
(1F), but each line has its own specific number of nodal diameters. Note how the upper
eigenfrequency lines are lying closer to each other. Since the disk becomes stiffer with increasing
number of ND, the eigenfrequency is approaching the blade alone frequency, as explained in
section 2.3 with figure 2.21. This means that the upper eigenfrequency line from figure 2.29
corresponds to one of the eigenfrequency lines (in this case the 1F line) in figure 2.28.
Adding the nodal lines to the Campbell diagram leads to crowdy and hard to use diagrams, see for
example figure 2.30 which is a combination of figures 2.28, 2.29.
2.7 Interference diagram
A better way to find dangerous rotational speeds of a bladed disk, is by representing the frequency
vs. the nodal diameters, [51]. This diagram is called an interference or SAFE (Singhs Advanced
Frequency Evaluation) diagram and was already used in figure 2.21 for showing the influence of
disk stiffness. A more extended diagram is drawn in figure 2.31 for a 64 bladed disk. Each line
represents the eigenfrequency of a bladed disk mode for all possible nodal diameters. The solid
lines are bladed disk modes with the blades vibrating in a blade mode, because they converge to a
blade alone frequency. The dotted line represents a bladed disk mode which is dominated by a disk
mode. For some modes it is harder to achieve a 0 ND mode than the 1 and 2 ND modes. That is
why most of the solid lines are characterized by a ―fish hook‖ at low nodal diameters. In figure 2.31
the nodal diameters are drawn up to N/2, but it is also possible to draw higher (indistinguishable,
see figure 2.19 in section 2.3) nodal diameters. This is done in figure 2.32 (dotted line) for one
family (= one eigenfrequency line from figure 2.31).
53
Figure 2.29 Campbell diagram for the first blade flapwise bending mode shape (at all possible nodal diameters) of a 36
bladed disk
Figure 2.30 Campbell diagram for 4 bladed disk mode shapes (at all possible nodal diameters) of a 36 bladed disk
54
Figure 2.31 Interference diagram for a 64 bladed disk
Figure 2.32 Interference diagram for one family mode of a 40 bladed disk
The solid lines, indicated with a rotational speed, in figure 2.32 represent constant speed lines. For
these lines the x-coordinate indicates the EO of possible excitation forces. Each point on the
constant speed lines is found by multiplying the EO(=ND) times the rotational speed (in rates per
second). Each intersection of the constant speed lines with an eigenfrequency line is a dangerous
excitation, because the excitation source and the blade disk have equal frequency and mode shape.
This diagram is drawn for 80 ND whereas 20 (= N/2) ND would be sufficient because higher nodal
diameter modes (> (N/2) ND) are indistinguishable. The intersections with the constant speed lines
can be retrieved by reflecting those lines on the vertical lines through the 0 and N/2 nodal diameter
coordinates. The intersections with the reflected speed lines (dash-dotted lines in figure 2.32) occur
55
at the same frequency as for the nonreflected speed lines, but one only has to calculate the bladed
disk modes up to N/2 ND.
It should be mentioned that the bladed disk mode (dotted line) in figure 2.32 is drawn for all
rotational speeds, but they lie on top of each other because the temperature and centrifugal (de-
)stiffening effects are not included. Including them would make the diagram rather complicated.
Since the temperature and centrifugal effects are counteracting each other, it is doubtable to include
them.
Next to the interblade coupling, the coupling between different mode shapes can be showed in
interference diagrams, see figure 2.33. Mode coupling occurs in zones where two or more
eigenfrequency lines of different bladed disk modes are close to each other. In these zones, called
veering zones, modes are mostly a combination of, for example veering 2 and 3, in-plane bending
mode shapes of the blades and out-of-plane motions of the disk, [49]. Taking a closer look to the
veering zones in figure 2.33, one can denote a difference between veering 1 and the other two.
Veering 1 has a wider nature (wider in frequency and nodal diameters), which indicates that there is
a strong coupling between the blade and the disk motions. The strong coupling between blade and
disk also means that the interblade coupling through the disk will be much stronger than at the
veering zones 2 and 3.
Figure 2.33 Mode coupling in interference diagram of a 12 bladed disk
56
2.8 Blade tip clearance variations
The accurate control of running gaps between static and rotating components is vital to preserve
mechanical integrity and ensure a correct functioning of any compact rotating machinery. In
modern aircraft propulsion systems, blade tip clearance plays a major role in the characterization
and development of reliable and efficient aero-engines. In current turbine stages, hot leakage jet
flows above the rotor airfoil tips (figure 2.34 b) resulting in 30% of the total aerodynamic losses
[52]. Moreover, during the machine operation the casing and the bladed disks undergo different
dilatations because of different thermal inertia values of the components. This result in blade tip gap
variations, as well as the shaft orbiting due to rotor eccentricity (figure 2.34 a). Therefore, the gap
between the rotor blade tip and the stationary shroud must be minimized. However, turbine
designers commonly set a safe minimum gap to prevent mechanical rubbing between the blade tips
and the casing that may result in rotor damage or catastrophic failure. Thus, the selection of an
optimum tip clearance results from a trade-off between aerodynamic performance and mechanical
integrity. Unfortunately, as a consequence of manufacturing and installation tolerances, vibrations,
different expansion rates of blades, rotor disk and casing induced by thermal and speed transients,
the rotor tip clearance undergoes large variations during engine service operations [1]. Hence,
advanced tip clearance active control concepts must rely on accurate real-time clearance
measurement systems to achieve highest engine efficiency with increased safety margins.
Additionally, blade tip clearance measurements are routinely used in engine development programs
to verify the mechanical design and tune numerical predictions [53].
In aero-engines the most challenging tip clearance measurements are carried out in the high
pressure turbine where temperatures, pressures and rotor speeds are the highest. Sensors should thus
be capable of withstanding temperatures above 1500 K, have a reduced size and improved working
range (0-5 mm), and a high bandwidth to resolve blade passing frequencies of the order of tens of
kHz. Tip gap measurements in current gas turbine applications are commonly based on capacitance
systems. Similarly to eddy current probes, capacitance-based tip clearance sensors are low cost and
rugged devices with the best potential to adapt to harsh working conditions. The technique relies on
the measurement of the capacitance between an electrode and a moving metallic target. Different
capacitive detection methods have been developed to sense the running tip capacitance. However,
no sensors for engine flight have yet been realized [4]. The gas turbine industry therefore has a
strong interest for improved tip clearance instrumentation for routine use on in-service engines.
57
Further technology development is needed to provide measurement systems with enhanced
accuracy, increased working range and extended lifetime.
Figure 2.34 a) Example of shaft orbiting b) Tip leakage flow
2.9 Conclusions
The aim of this chapter was to introduce two fundamental phenomena to monitor in
turbomachinery: the vibration of the disk and blades and the vatiation of the gap between the blade
tip and the casing. It is not trivial to monitor these parameters, especially by a contactless technique
and it is even more complicated to mitigate their effects on the machine. A clear understanding of
the theory behind vibration problems and clearance variations it is mandatory before analysing
measurement results.
58
3 Blade Tip Clearance and Blade Tip Timing measurement systems
3.1 Introduction
A simultaneous Blade Tip Clearance (BTC) and Blade Tip Timing (BTT) measurement systems is a
non-intrusive instrumentation for the measurement of blade tip clearance and blade vibrations. As
shown in figure 3.1, it is made of stationary sensors (S1, S2, …, Sn) mounted around the casing,
connected to dedicated electronics for the signal amplification and filtering, then to a data
acquisition system (DAS) and finally to a PC or a data processing and visualization system. BTC
and BTT systems are based on the analysis of key parameters of the sensors signals and time
intervals.
This chapter illustrates the principles behind the signal analysis of a BTC and BTT measurement
system.
Figure 3.1 Blade tip clearance and blade tip timing measurement system
3.2 The principle for the measurement of blade tip clearance
Whatever the sensor technology is, when a blade passes in front of a stationary probe an output
signal is generated. To monitor the clearance, the signal must be sensitive to the variations of the
gap between the head of the probe and the blade tip.
59
Figure 3.2 a) Typical sensor signal b) Influence of clearance variation
Every sensor type has its working principle and the shape of the output signal generated at the blade
passage is usually different. Nevertheless, the system identifies one or more key parameters
representative of the current blade tip gap. The key parameters must be robust and repeatable.
Typical key parameters used to monitor the clearance and its variation are the peak to peak value of
the output signal, the maximum peak value or the minimum. For instance, it can be considered a
typical output signal of figure 3.2a. The variation of the blade tip gap, in this case, induces a
variation of the signal amplitude, as in figure 3.2b. Hence, the key parameter in this case can be the
maximum peak value of the signal.
So far it seems easy. In reality, the key parameter can vary with external disturb like the machine
speed variations, different blade material, the temperature, etc.. It is not trivial to identify a key
parameter well representative of only the blade tip clearance and is even more difficult to
characterize and compensate for all the sources of uncertainties affecting the measurement chain.
Another important aspect to consider in the signal analysis is the speed of the process. Once the key
parameter is defined, there is somewhere in the system a signal processor that must extract this
parameter for every pulse, per blade, per revolution. Nowadays the DAS are very fast, but the
extraction of the parameter and conversion in tip gap (millimeters) must be fast and intelligent, to
avoid a bottleneck. Imagine to have a DAS with a sampling frequency of 1MHz per channel, to use
8 sensors around the casing and to have a rotor with 24 blades rotating at 10000 rpm. The system
must extract and convert 4000 maximum peaks values per blade, per revolution, per channel, in
real-time only for clearance measurements!
This is why, when designing the software to process the data, it is important to develop light
routines and intelligent procedure to cut the non-interesting parts of the signals.
60
3.3 The principle for the measurement of blade vibrations
BTT systems are based on complex processing methods developed for the first time more than 30
years ago from researchers of the Arnold Air Force Base, NASA and Brown Bovery [54-56].
This non-intrusive technique for the measurement of blade vibrations was born as substitute of
strain gauges measurements. The latter are intrusive and needs a complex telemetry system to
transmit the data. Moreover, BTT ensures the measurements of all the blades at the same time,
whereas the strain gauges are reliable and have a consolidate technique but referred only to these
blades where the gauge is applied.
The concept behind a BTT system is simple: when a blade is rotating and vibrating, the time of
arrival (TOA) of the blade tip in front of a stationary sensor will depend on the rotor speed and on
the displacement due to its vibration. Once the rotor speed is known, the theoretical time of arrival
can be predicted and the remaining time variations will only depend on blade vibrations. Hence,
blade vibrations can be measured by means of time variations.
To better explain this concept it can be considered the blisk of figure 3.3. There are 4 blades and 1
stationary BTT probe. Every time a blade passes in front of the sensor, a typical output signal (as
the one in figure 3.2a) is generated. During the first round, there are no vibration and 4 pulses
equally spaced in time are acquired. On round II, the blade 3 vibrates, the tip is backward deflected
when passes in front of the sensor and the corresponding signal is delayed (+ Δt) with respect to the
non-vibrating condition. On round III, the blade 3 vibrates with opposite phase, the tip is forward
deflected when facing the sensor and the corresponding pulse is in advance (-Δt) with respect to the
non-vibrating condition. By knowing the time variations (± Δt), it is possible to calculate the blade
displacements.
In case of no-vibration, the mounting angles of the blades are known (Δθ) and the TOA of every
blade will depend only on the rotor speed (3.1):
(3.1)
The disk speed (ω) is also known from a reference signal (REF), which can be a Once Per Rev
(OPR) sensor on the shaft or averaging the blade pulses on one revolution ( further detail on chapter
4).
61
In presence of vibration, the equation 3.1 is still used to calculate the theoretical TOA of every
blade per revolution. Nevertheless, in this case the real sensor signal will arrive shifted in time (±
Δt) due to the blade vibration. By measuring the time variations (± Δt), it is possible to calculate the
blade deflection samples Si:
(i: 1….N) (3.2)
where N is the number of blades and is the tip velocity, calculated as:
ω (3.3)
where R is the tip radius.
The motion of the blade tip is the superposition of two contributions: its rotation around the shaft
and its vibration. The analytical expression is:
( ) ω f ( ) f ( ) (3.4)
The first term of (3.4) is the contribution of the rotation, the second is the vibration. The latter
represent the generic formulation of the blade vibration to measure:
( ) f ( ) (3.5)
where f2(t) is a periodic function of period T, frequency (f = 1/T) and amplitude A.
Figure 3.3 The concept of the TOA variations
62
The idea of the BTT technique is to measure blade vibrations of generic formulation (3.5) from the
displacement samples (Si) of the blade tip measured from the stationary sensors. To do this,
complex algorithm calculates the TOAs of every blade from the REF signal, detects the key
parameter time variation (±Δt) from the sensor signal for every blade, converts in displacement
sample (Si), finally, manipulates and fits the data with different methods to retrieve the measured
vibration. Without going into detail, it can be imagine that more the measured samples, more the
fitting points, better the result. Nevertheless, more samples means more sensors, more electronics,
more data to process. Hence, the choice is a trade-off, between the costs, the complexity of the
system and the needed accuracy. (Further detail on the algorithm are presented on chapter 4).
The BTT systems measure blade vibrations which can be synchronous or asynchronous, depending
on the relation between the disk frequency and the blade vibration frequency. Synchronous
vibrations have a frequency which is an integer multiple of the rotational frequency. If this
condition is not satisfied, the vibration is asynchronous.
In case of a synchronous vibration (figure 3.4b), the measured displacement samples of one blade
will be the same at every revolution. It is the so-called ―stationary wave‖. One can realize that
having only one sensor it is not possible to measure vibrations in this condition. ( In reality this is
not fully true as it will be explained in the following chapters, nevertheless the accuracy of the
system is decreasing with the sensors number). Moreover, considering the asynchronous vibration
in figure 3.4a, it can be seen that, if the number of sensors is not enough, the fitting points to
―rebuild‖ the vibration are not sufficient and the system is working in aliasing condition.
An innovative approach to reduce the sensors number and improve their positioning around the
casing was presented by Rossi and Brouckaert [57]. To avoid aliasing, a minimum of 2 samples per
blade vibration period must be taken. Nevertheless, this can be not sufficient to correctly reconstruct
the vibration patch. Figure 3.5a shows a BTT system with 4 sensors every 90°. The black curve
represents the real blade vibration and the red dots the measured samples. In this first case, all the
sensors ―see‖ a zero displacement because they are placed in a wrong position, where the tip is not
moving. The work presented in [57] suggests to mount the sensors around the casing in a non-
equally spaced way, as in figure 3.5b. It is clear from the graph of the measured displacements that,
the samples now are well representative of the vibration patch. Finally, by having the sensor not-
equally spaced, it is possible to correctly reconstruct vibration patches at different frequencies
(figure 3.5c).
63
Figure 3.4 Example of a) asynchronous and b) synchronous vibration
Figure 3.5 a) Equally spaced sensors b) Non-equally spaced sensors c) Sampling of different frequencies with non-equally
spaced sensors
3.4 Conclusions
This chapter has introduced the basic principles behind a tip-timing and tip-clearance measurement
system. Although it seems easy, it requires very complex data processing algorithms and powerful
acquisition systems.
The aim of this research project was to investigate on the possibilities of using the magnetoresistive
effect on BTT and BTC systems. Hence, the basic principle here described were considered and
implemented in this project. Nevertheless, the system to develop in this phase could be based on
only one sensor and one channel processing method. Multichannels routines and dedicated
acquisition systems are already commercially available, therefore the innovation and the research
project is focused on developing a new sensor
64
4 Design of the hardware and the software of the measurement system
4.1 Introduction
This chapter illustrates step by step all the phases of the design and realization of the hardware and
software components of the measurement system.First the magnetoresistive elements are presented.
Then numerical and experimental tests for the sensor development are summarized. The third part
regards the first sensor prototype: the final assembly of the first prototype of the probe is reported.
Subsequently, the dedicated electronic for the signal amplification and filtering is described.
Finally, this chapter ends with the description of the universal calibration bench realized and the
fundamentals of BTT and BTC data processing.
4.2 The magnetoresistive sensing element
In 1857, William Thomson ( alternatively known as Lord Kelvin) discovered that the resistance of
the iron changes under magnetization. The property of a metal to change resistance with the
variation of an applied magnetic field is the magnetoresistive effect. At that time, one common
application of magnetic devices was the realization of power transformer. Some ferromagnetic
materials, under the influence of an applied magnetic field, present compression or expansion with
generation or loose of heat. This phenomenon is called magnetostrictivity. Around 1910 it was
discover a special alloy of nickel-iron (Ni81Fe19) with almost zero magnetostriction: the
permalloy. It is consider a soft magnetic material, in the sense that its magnetization can be easily
modified by external magnetic fields. Nowadays, magnetoresistive sensing elements are realized
with very thin film of permalloy on silicon wafers. The permalloy is an anisotropic material: its
magnetoresistivity depends on the direction of the applied magnetic field. In particular it has two
main axis: the Easy Axis (EA) which is the principal magnetic axis of the film and the Hard Axis
(HA), perpendicular to the previous one. Considering the film of permalloy of figure 4.1, its
resistance R can be calculate as follows [58]:
(4.1)
where is the resistivity of the material.
To better highlight the change of resistance with respect to its mean value, it is often used the
parameter percentage of variation of resistance, defined as:
65
Figure 4.1 Film of permalloy
⁄ (4.2)
In this project, the magnetoresistive sensing element Honeywell ―HMC1501‖ was used to build the
sensor. The Voltage supply of the sensing element is is 5 V, the angle of measure is from -45 to 45
deg, the sensitivity is 2 mV/deg, the measure resolution is 0.07 deg and the operating temperature
range is from -40 to 125 °C.
The HMC1501 is made of a saturated-mode Wheatstone bridge. The output voltage of the bridge is
function of the direction (θ) of the magnetic induction over the sensing element surface. The
change of the output signal with theta is presented in Figure 4.3; (further information in [58 – 60]
and in the appendix ).
Figure 4.2 The Honeywell HMC 1501
66
Figure 4.3 HMC1501 Signal output in function of theta
By placing the sensing element beside a permanent magnet and having a ferromagnetic object (like
a blade) passing in front of it, there will be a distortion of the magnetic field. This result in a change
of Theta, therefore in the sensor output. This concept is schematized in figure 4.4 and it was already
verified by Pavel Procházka [61]. In fact, figure 4.5 reports the output signals of a magnetoresistive
sensing element at the blade passage (figure 4.5 left) and in case of clearance variation (z) between
the blade tip and the sensor head (figure 4.5 right).
This is the only reference that has been found so far using Magnetoresistive sensors. He reported
blade tip clearance and blade tip vibration measurements performed in a steam turbine. Starting
from the state of the art on magnetoresistive probes, this project had the aim to further investigate
on the capabilities and the limits of the magnetoresistive elements in turbomachinery application.
The purpose was to better understand how to build a magnetoresistive probe for accurate and
repeatable measurements, improve its performance and realize a simultaneous real-time BTT and
BTC measurement system.
Figure 4.4 The sensor working principle
67
Figure 4.5 Left: Typical output value experimentally observed; Right: Sensor signal change with the clearance [41]
The basic configuration to realize a magnetoresistive sensor is made of the sensing element and a
permanent magnet. Anyway, moving the magnet around the sensing element, the output signal
changes. Moreover, using stronger magnet or placing the magnet with different polarization
directions, the output signal, for the same blade passing condition, vary as well. This is normal
because, modifying the magnet position and strength, the change in the direction of the magnetic
field seen by the sensing element is different.
These aspects gave reasons to realize an optimization of the sensor components layout. A simple
test bench was realized and the result are reported below. In the appendix F all the specifications are
collected.
4.3 Design of the sensor
To find the best layout of the components of the probe a test bench (figure 4.6) was realized. It has a
two blades rotor driven by an electric motor. A magnetoresistive sensing element HMC1501 was
placed on a fixed support in central position, facing the rotating blades. Behind, an x-y sliding
system was realized with two micrometric linear sliders OWIS MT 60-UM. It was used to place
different magnets in different positions around the sensing element. Test were conducted with two
magnets: ―S-04-25-N‖ (cylindrical shape) and ―W-05-G‖ (cubic shape).
For every position of the magnet, the signal corresponding to a blade passage was acquired. The
optimum layout is the one corresponding to the output signal with the highest amplitude.
68
The micrometric slider adopted are made of a non-metallic material, in order to do not affect the
magnetic field of the magnet. For the same reason, all the other components used are also made of
aluminium or plastic materials. Two sliders were used, mounted at 90° respect each others, to move
the permanent magnet on several points of a 2D investigation area. The grid of measurement points
is illustrated in Figure 4.7. The signals were stored with the Picoscope 4424 (4 channels, 12 bits
resolution, 20 MS/s, 32 MS buffer memory).
Figure 4.6 Test bench for the optimization of the sensor components layout
Figure 4.7 Top view of the bench and measurement grid
69
Measurements with the cylindrical magnet ―S-04-25-N‖
Table 4.1 lists all the peak to peak values (Voltage in [mV]) of the signals acquired at various
magnet positions. The coordinate system is the one displayed on the measurement grid of figure 4.7.
The x axis corresponds to a vertical displacement and the y axis to the horizontal direction. The
sensing element was placed at the coordinate (x;y) = (0;0) mm and the first measurement point was
taken with the magnet at (x;y) = (9;0) mm. Figure 4.8 shows all the peak to peak values in a contour
plot.
y axis
position [mm]
20 15 10 5 0
x ax
is
39 39,50 39,25 35,32 32,25 30,28
34 39,80 39,70 37,15 34,02 32,30
29 37,24 37,03 35,82 34,19 33,45
24 20,08 28,43 30,33 31,08 31,75
19 14,34 14,60 20,69 24,77 27,00
14 23,75 8,80 9,38 16,50 19,95
9 34,95 22,42 5,09 7,76 11,02
Table 4.1 Peak to peak output signal values
Figure 4.8 Peak to peak output signal values
70
The maximum output amplitude was obtained at the position (34,20) mm with a peak to peak value
of 39.8 mV. The data stored in Table 4.1 are displayed in three different colors because a change on
the signal shape has been also observed and details are reported in the appendix, part F.
Measurements with the cubic magnet ―W-05-G‖
All the measures were repeated with the second magnet. The output values have shown only small
changes. The maximum amplitude was obtained at position (18, 20) mm with a peak to peak value
of 36.1 mV.
4.4 The first sensor prototype
According to the results of paragraph 4.3, the best layout to realize the sensor is the one with the
cylindrical magnet ―S-04-25-N‖ placed at x = 34 mm and y = 20 mm from the sensing element. To
reduce the size of the probe, it was decided to have a symmetric layout with all the components on
the same axis y = 0. Therefore, the first magnetoresitive sensor prototype (figure 4.9) was built with
the cylindrical magnet ―S-04-25-N‖ placed at x = 29 mm from the sensing element.
It is made of an 8 mm outer diameter brass pipe, with the sensing element placed inside, at the top,
as it is possible to see from the image below. At 29 mm follows a permanent magnet connected
with a screw. The latter allows to change the relative distance between the magnet and the sensing
element, in case of necessity.
Figure 4.9 The first magnetoresistive sensor prototype
4.5 The mathematical model of the sensor
During the sensor design phase several tests were performed to find the best probe configuration.
Hence a large number of experimental data was collected. Therefore, it was developed a
mathematical model of the probe and validated with the experimental results. The big advantage of
71
a FEM model is to have a reference when trying to develop new sensor configurations or investigate
on new applications.
The magnetic field is generated by the NdFeB permanent magnet in shape of a rod having diameter
4 mm, height 12,5 mm, weight 1,2 g, Curie temperature 310 °C, magnetic remanence from 1.29 to
1.32 T, coercive magnetic field strength from 860 to 955 kA/m. The direction of the magnetic
induction changes when the position of the ferromagnetic iron blades changes. If the sensor bridge
is placed in a position where the change of the angle of the magnetic induction is relevant, it gives a
sensitive output variation and small noise, resulting in high Signal to Noise ratio.
The FEM analysis of the prototype was performed using the numerical code Ansys Maxwell Rel
14.5. It was modeled the 3D geometry of the compressor blade mounted on the test bench shown in
figure 4.6, and the permanent magnet described above. These objects are included in an air cube
where the boundary conditions have been imposed along its outer part. In particular it was imposed
the parallel flux condition. The air cube edge is ten times longer than the length of the blade. It was
seen that for larger boundary region the numerical results do not change anymore. The problem
geometry has been discretized using about 18,000 tetrahedral elements. In figure 4.10 is shown a
particular of the blade tips volumes designed to take into account the different blade positions
respect to the permanent magnet position. For each blade position the corresponding volume is
defined as steel, while the others volumes are defined as air. The average relative magnetic
permeability of the blade steel has been assumed equal to 120. This value has been deduced by
static measurements performed in laboratory [62-64]. In the volume of the permanent magnets the
material parameters used were the residual magnetic induction and the coercive magnetic field. The
permanent magnet has been paced at d = 29 mm respect to the blade tip as schematized in figure
4.10. The magnetoresistive sensors can be placed at different positions between the compressor
blade and the permanent magnet. In order to optimize the sensors number and their positions it was
computed the three components of the magnetic induction in 1000 points along the distance d. This
simulation has been repeated for 21 values of the blade angle from -20 degrees to +20 degrees
with a step of 2 degrees. In this way we have generated a map of the magnetic induction change
when the blade moves, rotating in front of the permanent magnet. This map is shown in the figure
4.11. The magnetic induction change is represented by the so called magnetic induction deflection,
that is the variation of the angle of the magnetic induction vector M during the blade rotation. This
deflection is related to the x-y plane, and the deflection respect the z axis is negligible. The results
presented in the figures confirm that the higher sensitivity can be obtained using a sensor positioned
72
in proximity of the blade tip when it faces the permanent magnet. More precisely, in figure 4.11 left
is reported a comparison between experimental and numerical data of the magnetic distortion at the
blade passage.
Figure 4.10 Typical modelling FEM mesh at different blade positions
Figure 4.11 The magnetic induction change at the blade passage
4.6 The basic electronic board
A preliminary electronic board was developed following the electronic circuit suggested by
Honeywell in the datasheet [60], as in figure 4.13.
It correctly works for static measurements and low dynamic conditions. Nevertheless, when trying
to use magnetoresistive sensors in high speed turbomachines with high blade passing frequency, it
is highly recommended to develop a dedicated electronic.
In this work several modifications were made to the basic electronic board in order to avoid
bandwith limitations and add amplification, filtering and signal conditioning stages.
73
Figure 4.12 Basic electronic circuit for static measurements
4.7 The Universal Calibration Bench for BTT & BTC systems
A light and handy test bench was manufactured by the University of Perugia for the characterization
of sensors for rotating components [67]. As it is possible to see in figure 4.13 it is composed of an
electric motor that drive a rotating disk of diameter 100 mm having four small plates of steel
(thickness 1 mm) every 90°. It is used to simulate the blades passages up to 25000 rpm. Around the
disk there is an alluminium casing for safety reason. On the top of the casing, the sensor housing is
installed on a vertical linear slider fixed to the frame. In this way, it is possible to vary the gap
between the sensor head and the tip of the thin steel plate.
The first version of the test bench was upgraded according to the idea of the blade motion
decomposition published in [57]. It was recently reported by Tomassini et al [68]. Instead of having
fixed casing sensors and vibrating blades, the innovative approach consist in letting the sensor
vibrate while non-vibrating blades are kept in rotation. The biggest advantages of this system are
the possibility to know the expected measured vibrations and the flexibility in changing the test
conditions. It represent a Universal Calibration Bench for any type of sensors used in BTC & BTT
systems.
First of all, a new disk was designed. The latter is provided of 24 symmetrically distributed holes to
mount properly shaped and ―rigid - blades‖, fixed at the disk with screws (figure 4.14).
74
Figure 4.13 The test bench for the sensor characterization – version 1
Figure 4.14 The new disk and the little blade
Figure 4.15 The bench equipped with a sensor shaking system
75
Subsequently, in spite of having a vibrating blade and a fix sensor to acquire the blade passing
pulse, the idea is to let the sensor vibrate at a known frequency and amplitude, meanwhile the rotor
with the little blades, supposed to do not vibrate, is spinning. To impose a known vibration to the
probe, the test bench was equipped with a shaker, as in figure 4.15. The linear slider where the
sensor is mounted is connected to the shaker and a signal generator makes it possible to impose a
vibration of a certain frequency and amplitude. The maximum achievable displacement at different
frequencies has been experimentally obtained by means of an optical displacement measurement
system (Appendix E, Figure E1). The vibrating system can achieve a maximum displacement of
about 2.3mm up to 30Hz, than the mass of the moving parts become relevant and the system
response decrease with the square of the frequency.
All the components were grouped together with a compact frame, resulting in a Universal
Calibration Bench for any type of sensor for BTC & BTT systems. As shown in figure 4.16 and
4.17, the facility is made of: a data acquisition system (DAS), the electronic board, a signal
generator connected to the amplifier of the shaker, the latter fixed to the moving sensor housing
placed at the top of the test bench.
Figure 4.16 Left: the frame. Right: the assembly of the universal calibration bench
The displacement of the sensor under calibration is monitored by a displacement probe with high
resolution mounted on the fixed frame. It was calibrated with an external optical system (figure
4.15) with a resolution of 3 µm.
76
Figure 4.17 The Universal Calibration Bench for BTC & BTT systems
Figure 4.18 The displacement monitoring sensor of the shaking system
4.8 Design of the post-processing software
As presented in chapter 3, a BTC & BTT software must extract the key parameters from the sensor
signals to give information on the current blade tip gap and the presence of blade vibrations. These
routines already exist and are commercially available. The data processing in tip-timing is complex,
is directly affecting the results and is very difficult to find references because every company has its
method and it keeps it secret. Nevertheless, this aspect can still be improved developing new
methods of data processing.
77
In the frame of this research project, it was developed a post-processing software in Matlab for the
analysis of the magnetoresistive signal. It should be considered that the aim was not to improve the
existing professional and real-time routines, but to have a robust data processor to evaluate the
performance of the innovative magnetoresistive sensors for turbomachinery testing. Therefore,
below the main phases of the algorithm developed are only summarized to give an insight to the
readers.
The algorithm is divided in three main parts: reading, calibration and processing.
The first phase opens the signal, subdivides it in revolutions, acquires the settings of fundamental
parameters like the sampling frequency, the rotor radius, the number of blades, etc.. The calibration
routines check if the signal has errors. In fact, it happens that there is a missing blade, the signal is
noisy and requires a smoothing procedure, etc.. Finally, the processing routines are the core of the
software.
The processing procedure is divided in seven sub-routines:
- Calculation of the maximum peak values
- Calculation of the zeros
- Interpolation
- Integration
- Extrapolation of the TTOAs
- Calculation of the blade displacement samples
- Plotting of the results
1.Calculation of the maximum peak values
A typical output signal corresponding to a blade passage was shown in figure 4.5. Concerning the
clearance measurements, the software extracts a key parameter (in this case the maximum peak
value) for every pulse generated by every blade passage during the acquisition. Figure 4.19 shows a
typical signal collected in the lathe and representing few rotations of a gear at low speed (blue line).
Sometime there is also a reference signal, one per revolution (1/Rev or OPR), here represented by
the red line.
78
Figure 4.19 Typical signal collected with the sensor prorotype and a gear rotating in the lathe
Once a threshold has been defined by the user, the program split the whole signal in the upper and
the lower part. At that time, it retrieves each max value, for every peak of every blade, as in figure
4.20. When the calibration curve of the sensor is known, is possible to convert the measured volts in
millimiters of blade tip gap.
Figure 4.20 Detail of the calculated maximum peak values
79
2.Calculation of the zeros
As introduced in chapter 3, the fundamental parameter to extract from a tip-timing signal, is the
blade time of arrival (TOA). Having signals as figure 4.5 right, it can be noticed that, whatever is
the clearance, all the pulses passes through the zero voltage point. This point is not affected by
clearance variations and it is a good representative of the TOA. Figure 4.21 illustrates a typical raw
signal and the relative TOA for every blade. The signal around the zero voltage value is
reconstructed by a fitting curve to avoid the noise and to increase the calculation accuracy.
Figure 4.21 Example of TOA calculation
3.Interpolation
Once the TOAs of one revolution are known, the next step is the calculation of the Δt between each
TOA. The values Δti in one revolution and the mounting angles of the blades Δθi are used to
calculate the samples of the angular velocity (ωi) over one revolution, as follows:
i = 1 ….N (where N is the number of blades) (4.2)
The Δti and the ωi are fi ed wi h a weigh ed polynomial o know he rend of ω in fun ion of
he ime (ω( )).
4.Integration
The function ω(t) is integrated over the period of one revolution to obtain the angular position of
the blades in function of the time: θ(t).
80
5.Extrapolation of the TTOA
At this point, it is possible to extrapolate the theoretical time of arrival (TTOA) of the incoming
blade. For instance, TTOA1_2 will be the expected time of arrival of blade 1 at the revolution 2.
6.Calculation of the blade displacement sample (Si)
The incoming blade is expected to arrive at the TTOA extrapolated. Its real TOA will be calculated
and equal to the time corresponding to the zero crossing voltage. Then, the blade displacement
sample Si will be:
{ (in a e of no i ra ion )
( ) ω (in a e of i ra ion )
(4.3)
For every incoming blade signal, the ω(t) is upgraded of one point, it is newly integrated, the new
TTOA is extrapolated and the next blade displacement sample Si is calculated. This iteration is
schematized in figure 4.22. It stops at the end of the acquisition. Finally the results are plotted.
Figure 4.22 Scheme of the TTOA and TOA calculation
81
7. Plotting of the results
The last part of the algorithm regards the plotting of the results. The software plots the individual
blade displacement measured for every blade and the waterfall, which is an ensemble plot of the
displacement of every blade. Figure 4.23 and 4.24 shows two example of plotting of results which
will be explained in the following chapters.
Figure 4.23 Example of measured blade displacements
Figure 4.24 Example of waterfall of the measured blades displacements
82
4.9 Conclusions
Several problems were fixed while developing and testing the system. This chapter summarized the
design phase of the different components of the system and the following conclusions can be
drawn:
The experimental investigations have shown that, by changing the magnet position, big
variation results both in the signal shape and in the peak to peak value. A database of
different possible configurations was stored. Varying the magnet strength instead, the output
signal do not change so much.
Concerning the dedicated electronics, to avoid bandwidth limitations it is important to find
the right compromise for the gain of the system.
83
5 Blade tip clearance measurements
5.1 Introduction
The importance of monitoring the blade tip gaps and the fundamental parameters involved were
described in chapter 2 and 3. In this chapter the characterization of the magnetoresistive sensor and
its calibration curve for clearance measurements are reported. The performance of the probe were
evaluated on a static test at the blade passage, then a preliminary test was performed in a lathe using
a gear. Finally, the calibration curve of the magnetoresistive sensor for clearance measurements was
obtained in the Von Karman R2 compressor rig.
5.2 Characterization of the magnetoresistive sensor at the blade passage
5.2.1 Test setup and methodology
The passage of a blade in front of the sensor at different gaps was studied in a static test. A shaft
with two compressor blades was mounted in a rotary slider with a resolution of 1°, facing the
magnetoresistive sensor installed in a linear slider with a resolution of 10 µm. The sensor axis was
aligned with the blade axis at 0° on the rotary slider. A scheme is illustrated in Figure 5.1 left. The
blade passage was discretized and studied in the angular range [-20 + 20]° with an angular step of
2°. The measurements were repeated for different gaps (sensor head – blade tip): [0.5, 1, 1.5] mm.
At every fixed position, the output DC signal was acquired and represents one point of the resulting
fitting curves of Figure 5.1 right.
5.2.2 Results
Figure 5.1 shows the typical pulses observed at the blade passage at different gaps. The variation of
the clearance results in a signal amplitude variation: the bigger the gap, the smaller the signal
amplitude. Therefore, the clearance between a passing blade tip and the sensor head can be
monitored by the maximum peak value or the peak to peak value of the output signal. The latter is
not symmetric, most probably because of asymmetries in the sensor prototype or in the blade shape.
Nevertheless the peak to peak value can be used to measure the blade tip clearance and the
asymmetries in the sensor can be avoided with a product engineering.
84
Figure 5.1 Left: scheme of the clearance test. Right: blade passage pulse at different clearances
5.3 Preliminary clearance test with a gear
5.3.1 Test setup and methodology
A preliminary test was performed using a metallic gear, with 12 teeth mounted in the lathe. The aim
of the experimental investigation was to verify the repeatability of the measurements at different
velocities. The magnetoresistive sensor (S1, in Figure 5.2) was placed in the moving tower of the
machine and a second sensor (S2), used as OPR signal. The signals of one revolution were
acquired at three different gaps between the sensor head and the tooth tip and at different velocities.
In particular, the acquisitions were performed at: gap 0.1mm at 45rpm, gap 0.3mm at 500rpm and
gap 0.5mm at 1000 rpm. A comparator was used as displacement monitoring system (resolution of
0.01mm). The DAS used was The Picoscope 4424 (4 channels, 12 bits resolution, 20 MS/s, 32 MS
buffer memory).
Figure 5.2 Setup of the preliminary clearance test
85
5.3.2 Results
In figure 5.3, there are all the 12 pulses of one revolution plotted together at the different gaps and
velocities. First of all, to note that every blade gives a different output amplitude, most probably
because of its magnetization feature. Nevertheless, the variation of the maximum peak values with
the gap, listed in table 5.1 and plotted in figure 5.5, is almost the same despite different absolute
values. Moreover there was no significant variation of the output signal with the velocity, in the
speed range of the test.
Figure 5.3 Acquired signals at different gaps and velocities
Figure 5.4 Detail of the peaks of tooth 3, 4 and 5
86
max peak values
Gap 0.1mm Gap 0.3mm Gap 0.5mm
Peak 3 5.26 4.7 4.23
Peak 4 3.95 3.44 2.98
Peak 5 4.39 3.81 3.31
Table 5.1 Maximum peak values variation
Figure 5.5 Maximum peak values variation
5.4 Clearance test in the VKI R2 compressor rig
5.4.1 Test setup and methodology
Considering that the final application of the sensor objective of this research project are clearance
and vibration measurements in aero engines, the probe was finally tested and calibrated in a real
turbomachine at the von Karman Institute in the VKI R2 test rig (figure 5.6). It is an high speed
compressor driven by a 185kW DC motor. The rotational speed can vary up to 10000 rpm. It is a
single stage axial compressor with a 400 mm diameter test section. The rotor is composed of 24
subsonic blades of the NACA 65 family with the hub radius of 100 mm and the tip radius of 200
mm. The stator is composed of 30 blades. Four circumferential locations, every 90°, allow the
insertion of probes. Further details of the facility can be found in the appendix A.
The magnetoresistive sensor was installed in the calibration tower of Figure 5.7. It is equipped with
a comparator to monitor the displacement of the sensor (resolution 0,01mm) and it is mounted in
87
one of the four instrumentation access windows of the compressor casing. The passage of one of the
24 rotor blades was extracted and studied for three velocities v = [1000 - 3000 - 5000] rpm and for
different gaps in the range g = [1 – 1.5] mm with a step of 0.1mm. The gaps were changed moving
the calibration tower. For every combination velocity-gap (v,g), the machine was kept at fixed
speed, the sensor held at the desired gap and 30 revolutions were acquired.
Figure 5.6 The VKI R2 compressor rig
5.4.2 The calibration curve of the magnetoresistive sensor for clearance measurements
The mean value among the 30 maximum blade signals was extracted and it corresponds to
―Vmax(v,g)‖. For instance, Vmax(1000,1.2) is the mean maximum peak value acquired at 1000 rpm
and 1.2 mm clearance. The Vmax values were normalized according to (5.1). The resulting points
were plotted together and fitted with a linear function. Hence, in figure 5.8 is shown the calibration
curve of the sensor for clearance measurements. As example, it also shows the 30 maximum peak
values acquired at 1000 rpm and 1.2mm gap.
( )
(5.1)
Where:
Vmax(v,g) = the maximum peak value, for the combination velocity-gap (v,g);
Vmax = the maximum among all the values Vmax(v,g);
Vmin = the minimum among all the values Vmax(v,g);
88
Figure 5.7 The calibration tower of the R2 compressor
Figure 5.8 The calibration curve of the magnetoresistive sensor for clearance measurements
89
5.4.3 Uncertainty analysis
The expanded uncertainty U95 is defined as (for further details see [69] and appendix D ):
[ ]
(5.2)
Where:
b = the systematic component of the uncertainty;
SX = the standard deviation of the calibration data, defined as (3) – [SX = 0.0096 mm];
∑ ( )
(5.3)
Where:
Xi = the ith data point used to calculate the calibration line fit;
= the average of the calibration data;
N = the number of data points used to calculate SX - [6 gaps, 3 velocities = 18 points];
K = the number of curve fit coefficients [K=1];
In this case, the systematic component of the uncertainty was given a uniform distribution.
Therefore, the value of b was taken as the comparator resolution (0.01mm) divided by the root
mean square of 3 [70]. The resulting expanded uncertainty is: U95 = ± 22 µm (20:1) within the range
[1 ÷ 1.5] mm of gap range.
5.4.4 Results
A good repeatability was seen in the static and dynamic tests reported. The blade tip clearance can
be measured by means of the maximum value of the sensor signal peak. No significant dynamic
effects were observed. Moreover, the accuracy of the new magnetoresistive probe is in line with the
main types of sensors often used in BTC systems [71].
90
Figure 5.9 Detail of blade signals at different gaps
5.5 Conclusions
In the speed range tested so far, there was no signal change with the speed. Instead, what is
important to consider is the change of the signal amplitude with different blades. Measurements of
clearance are possible, but the peak to peak value of the signal depends on the blade and a
normalization is needed. This limitation can be avoided introducing a new configuration of the
sensor, as it will be described in chapter 7.
Accurate clearance measurement are possible with an uncertainty of ±22 μm within the range [1 ÷
1.5] mm of gap variation. The accuracy is very similar of what is possible to obtain with capacitive
probes. This point out very promising performance for the application in BTC systems.
91
6 Blade tip timing measurements
6.1 Introduction
This chapter focuses on the most relevant tip-timing test performed during this project. First, a
simulation of synchronous and asynchronous vibration in the universal calibration bench is
reported. Subsequently, two external collaboration with Fogale Nanotec (Nimes, France) and the
ITWL (Warsaw, Poland) has given an important contribution in improving the instrumentation.
Results of the external tests are illustrated. Finally, the optimized measurement system was installed
in the VKI R2 compressor rig.
6.2 Simulations of synchronous and asynchronous vibrations
The components of the measurement system (sensors, electronics and software) were realized during the on-
going collaboration between VKI and the Italian Universities of Perugia and Padova. A large part of the
development process was performed in the innovative universal calibration bench (paragraph 4.7). The most
significant test performed in this facility regards a simulation of an asynchronous and a synchronous
resonance.
6.2.1 Test setup and methodology
Tip-Timing signals have been acquired along a speed transient from 2000 rpm up to 15000 rpm.
The frequency of the imposed vibration to the sensor (by means of the shaker) was kept fixed at
100Hz, from the beginning to the end of the test. The amplitude was kept constant at 0.1 mm up to
13000 rpm, where it was decreased to about 0.05 mm. Moreover, two excitations were simulated
increasing the amplitude of the excitation up to 0.2mm. The first simulated resonance was at 4000
rpm, the second at 6000 rpm. The Campbell diagram of the test is shown in Figure 6.1: the fixed
excitation at 100Hz is represented by the red line. At 4000 rpm it does not cross any engine order
lines (EO), the excitation is not an integer multiple of the revolution frequency and it represents an
asynchronous vibration. At 6000 rpm, the red line crosses the 1EO line, in fact, the frequency of the
imposed vibration is 1 time the revolution frequency and it correspond to a synchronous vibration.
92
Figure 6.1 The Campbell diagram of the test
6.2.2 Results
Figure 6.2 shows the waterfall of the measured displacements for the first seven blades. The two
resonances at 4000 rpm and 6000 rpm are clearly visible. At 6000 rpm the measured displacements
show a change in the DC component of the signal, as typical for synchronous vibrations. The
presence of the two simulated resonances is evident, even though the constant vibration level at
0.1mm is always present up to 13000 rpm.
From the waterfall, the Individual Measured Displacements (IMD) of blade 1 (Figure 6.3), were
processed further. A window of 200 samples was defined and successive Fast Fourier Transforms
(FFTs) were performed, with a step of 10 samples. The 3D individual blade spectrogram of Figure
6.4 was obtained. Having only one vibrating sensor, one blade displacement sample per revolution
is measured. In this way, the sampling frequency (SF) of the IMD is the revolution frequency and it
is changing with the speed during the transient. Therefore, the vibration at 100 Hz measured in the
IMD is aliased up to 12000 rpm and it is folded at different frequencies, depending on the SF (to
note that SF = speed). This results in a zig-zag patch in the spectrogram [72], up to 12000rpm. As
expected, there are visible peaks around 4000rpm and 6000rpm. Finally, the spectrogram of Figure
6.4 shows the zig-zag patches of the harmonics of 100Hz and electric noise at 50Hz. These results
are even clearer in Figure 6.5, which represents the 3D spectrogram seen from the top, with the
amplitudes in color: the so-called Zplot.
93
Figure 6.2 The waterfall of the measured displacements of the first seven blades
Figure 6.3 Individual measured displacements of blade1
Figure 6.4 3D Spectrogram of the individual measured displacements of blade1
94
Figure 6.5 The Zplot of the measured displacements of blade 1
6.3 External validation test at Fogale Nanotech
Fogale nanotech is a manufacturer of BTC and BTT systems based on capacitive probes. It has 30
years of expertise in the field of capacitive metrology. During this research project an external test
of the developed instrumentation was performed at Fogale aiming at validating and improving the
system.
6.3.1 Test setup and methodology
Six Fogale capacitive probes and two magnetoresistive sensors were installed in the test bench of
Fogale (figure 6.6). Simultaneous tip-timing measurements of the rotating wheel were acquired and
processed with the Fogale software. The details of the test bench and the acquisition system are
confidential.
95
Figure 6.6 The Fogale test bench
6.3.2 Results
The results of the data processing with the two different type of sensors are in strong agreement and
the resonances were correctly measured. No problems or missing blades were encountered. The
magnetoresistive signal was compatible with the processing system and the whole instrumentation
developed was successfully validated. The acquired signals with both the sensors were very useful
as test case to improve the dedicated software developed so far. Below some real-time results were
saved during the acquisition. Figure 6.7 and 6.8 shows the waterfall of the measured displacements.
It can be seen, from the change in the DC component of the signals, that both the sensors were
measuring a resonance around 8000rpm. Finally, figure 6.9 and 6.10 reports the Zplot obtained at
the end of the test. At that time a FEM analysis of the rotating wheel was not available. Hence, the
folded frequencies in the spectrum are unknown, but it can be seen that the zig-zag patches strongly
agree meaning the measurement of the same phenomena. Quantitative comparison were not
possible, but the signals acquired helped in fixing software bugs.
96
Figure 6.7 A detail of the waterfall of the measured displacements with the capacitive sensors at Fogale
Figure 6.8 A detail of the waterfall of the measured displacements with the magnetoresistive sensors at Fogale
97
Figure 6.9 The Zplot from the measured displacements with the capacitive sensors at Fogale
Figure 6.10 The Zplot from the measured displacements with the magnetoresistive sensors at Fogale
98
6.4 External validation test at ITWL
The ITWL in Warsaw, Poland, is a military research institute of technology working with real aeroengines and
developing its own instrumentation. This second external test was performed to try the magnetoresistive
sensors on a real aeroengine, in order to have a real test case as reference and to compare the results with a
different measurement system.
6.4.1 Test setup and methodology
The magnetoresistive sensors were mounted in the military SO-3 turbojet engine of Figure 6.11. Four sensors
were installed in slots prepared in the upper compressor casing in the plane of the second rotor stage of the
compressor (Figure 6.11c). The rotor has 41 blades and a tip radius of 207 mm. Tip-timing data were acquired
during a speed transient up to 16000 rpm. A piece of felt was attached to the Inlet Guided Vanes (IGVs) to
excite the resonances. Previous tests performed at ITWL showed two blade resonances around 8000 rpm and
12000 rpm. The objective of the test was to repeat the experiment to validate the diagnostic system based on
magnetoresistive elements and collect reference data.
Figure 6.11 (a-b) The SO-3 turbojet engine – ITWL, (c) The 4 slots in the casing to install the sensors
6.4.2 Results
The magnetoresistive signals were recorded and processed by the ITWL Tip-Timing software. The
waterfall of the measured displacements for the first 15 blades are shown in figure 6.12 for the
ITWL eddy current sensors and in figure 6.13 for the magnetoresistive sensors. Both the waterfalls
show the two expected resonances around 8000 rpm and 12000 rpm. Some blades were exited at
slightly different rpm, which can be due to mistuning. From the FEM analysis on the blade
geometry performed at ITWL, two blade resonances were expected around 8000 rpm and 12000
99
rpm. In particular, the first mode is excited by the EO3 at 12000 rpm and by the EO4 around 8000
rpm. The data could not be processed further to have an estimate of the measured vibration
frequencies. Anyway, it is possible to compare the results at the stage of the measured
displacements. Figure 6.14 focuses on the measured displacements around 12000 rpm. Figure 6.15
and 6.17 are obtained by overlapping the measured displacements with the different sensors around
the second resonance, for blade 1 and blade 6. The peaks of the measured resonances in the two
tests are in strong agreement, both in amplitude and with respect to the speed. The expected
resonances were measured correctly and this confirm the possibility of performing tip-timing
measurements with the novel magnetoresistive probes. To notice that the measurements performed
with the magnetoresistive sensors show a drift. There might be several reasons. Further
investigations on the single components of the measurement chain will improve the system, anyway
the drift did not compromise the results. Finally, the whole instrumentation chain was successfully
working in a real engine for the first time. The magnetoresistive sensors were compatible with the
ITWL acquisition system and processing software. No damage was reported. The dedicated
electronic boards developed at VKI were successfully working up to 11 kHz blade passing
frequency and no particular noise problems were observed.
Figure 6.12 The waterfall of the measured displacements of the first 15 blades, ITWL sensors
100
Figure 6.13 The waterfall of the measured displacements of the first 15 blades, magnetoresistive sensors
Figure 6.14 Comparison of the measured displacements around 12000 rpm
Figure 6.15 Comparison of the measured displacements around 12000 rpm, blade 1
-0.4
-0.2
0
0.2
0.4
0.6
11000 11500 12000 12500 13000
Dis
pla
cem
en
t [m
m]
Speed [rpm]
MAG_pk1
ITWL_pk1
101
Figure 6.16 Comparison of the measured displacements around 12000 rpm, blade 6
6.5 The VKI R2 compressor
The validated and improved instrumentation was installed in the VKI R2 compressor (see paragraph
5.4.1) to acquire tip-timing signals during an acceleration transient. There are no FEM models or
references on the modal analysis of the rotor. Therefore, the characterization of the blades was
performed experimentally.
6.5.1 Test setup and methodology
Three sensors were mounted in the casing at [0 90 180]°. Tip-timing signals were acquired along a
speed transient from 2000 rpm up to 5500 rpm, with a sampling frequency of 1 MHz per channel.
No information was available regarding the expected blade vibration modes and frequencies. For
this reason, the hammer test with an accelerometer was performed on the blades to have an estimate
of the first bending frequency. It was assessed to be around 480Hz. Considering this and the speed
range of the test, it is possible to get the Campbell diagram of figure 6.17. The red line represents
the first bending frequency of the blades. It should account for the contribution of the centrifugal
force at different rpm, which increases the blade resonance frequency with the speed, but it could
not be estimated at this stage. The Engine Order lines (EO) indicate the multiple integers of the
rotor rotational frequency. For instance, the 2EO is the line fitting all the points equal to 2 times the
rotational frequency. When a line representing a blade vibration mode crosses an EO line, a
synchronous resonance can be excited at the crossing speed. Therefore, during this test,
synchronous blade resonances may appear around 2800, 3200, 3600, 4200, 4800 rpm. To let the
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
11000 11500 12000 12500 13000
Dis
pla
cem
en
t [m
m]
Speed [rpm]
MAG_pk6
ITWL_pk6
102
compressor works in off-design condition, a distortion grid was mounted at the rotor inlet (Figure
6.17 right). Moreover, an air injection system at the blade tip was installed to excite asynchronous
resonances. Four bent pipes were installed in the casing, two at 0° and two at 180° (figure 6.17
right). Compressed air at 5 bar was blown against the blade tips at about 3500 and 5400 rpm.
Figure 6.17 Left: the VKI R2 Campbell diagram. Right: the compressor inlet
6.5.2 Results
The clearance variation during the test was negligible; nevertheless the magnetic distortion induced
by the blades of the R2 compressor was significant, resulting in high signal to noise ratio.
Concerning the measurement of the vibrations, figure 6.18 and 6.19 show the waterfall of the
measured displacements of the first 4 blades. The measured displacements of blade 2, 3 and 4 were
shifted (by adding a DC constant value) to have an ensemble graph. The synchronous blade
resonances appear in the waterfall as a change in the DC component and the maximum peak value
of this change is taken as representative of the vibration amplitude. The asynchronous blade
resonances appear as a change of frequency with respect to the base frequency of the noise of the
system and the maximum peak value of this change is taken as representative of the vibration
amplitude. As shown in Figure 6.18, a synchronous resonance was measured around 3650 rpm and
an asynchronous resonance around 3450 rpm, as expected. Nevertheless, the measured amplitudes
are moderate. More evident is the synchronous resonance measured around 4900 rpm (Figure 6.19).
The latter was expected to be around 4800 rpm, but the effect of the centrifugal force was not
considered. Finally, the asynchronous resonance around 5400 rpm is also present.
103
Figure 6.18 The measured displacements in the R2 compressor, part 1
Figure 6.19 The measured displacements in the R2 compressor, part 2
6.6 Conclusions
Measurements of blade vibrations were carried out on test benches and real turbomachines with
very satisfactory results.The universal calibration bench realized was very important for the
development of the whole instrumentation. There are still margin of improvement, but comparisons
with different sensors and simulations of vibration resonances can be easily handled. Performance
of the magnetoresistive BTT and BTC systems were improved step by step in different test cases.
The developed instrumentation always resulted successful and very promising. No missing
resonances were observed and the dedicated components always worked well.
One aspect to further investigate regards the drift on the sensor signal appeared on the
measurements in the real jet-engine. It is not compromising the results, but further studies can avoid
this disturb.
104
7 Developments
7.1 Introduction
During this research project the whole instrumentation developed had several improvements. Bugs
and errors were corrected in the software, bandwidth limitation avoided in the electronics and
different sensor configurations were tested. This chapter reports the main developments achieved so
far. First, a real time software based on the post-processing routines described in paragraph 4.8, was
realized. Then, a differential sensor which avoid the calibration process, is illustrated. Finally,
preliminary test were performed in realizing an anti-aliasing software for tip timing data.
7.2 The real-time software
The algorithm described in paragraph 4.8 was developed and improved during the 3 years of this
research project. From the first configurations, several bugs and errors were corrected and it
becomes a robust software. Nevertheless, it is a post-processing algorithm which means that the
signals are processed after the acquisition and the results came after the processing procedures.
Hence, it was re-designed in GCode in Labview to implement the same logic but in a real-time
processing environment. Parts of the calculation routines developed in Matlab were re-used and the
measured displacements are always the same. This paragraph illustrates the acquisition system used
and some details of the GCode in Labview.
7.2.1 Tip-Timing and Tip Clearance data acquisition in streaming
The realization of a real-time software for BTC and BTT data analysing requires very powerful
acquisition systems, with high sampling frequency and several channels. In other words: performing
and expensive components.
In this project, only one channel was used in all the acquisitions. The routines developed are not at
multichannel level, therefore the real – time algorithm could be implemented with a USB Picoscope
5444B. It is a USB acquisition board with:
4 channels;
Resolution 8/16 bit;
Bandwidth up to 200 Mhz;
105
512 MS of buffer;
PC interface USB 2.0 high-speed;
Internal signal generator.
In real-time processing, the acquisition hardware can work in streaming or in block mode. The
first one allow the continuous acquisition of the signals from the channels and the communication
with the software of processing. The block mode is a faster communication, but it has a time
dedicated to the acquisition and a time dedicated to the transfer of data to the pc in which the
system is blind for the signal.
In this case, it is not acceptable to interrupt the signal acquisition and to lose data, therefore the
Picoscope was configured for streaming mode. The maximum sampling rate achievable in this
mode is 200 KS/s, with 8 bit of resolution. The hardware needs special setting to configure prior of
the acquisition. The manufacturer, on the website, provides the library at this scope:
ps5000Open to open the communication between the software and the Pico;
ps5000Settings to setup the channels and other parameters;
ps5000stream to setup the streaming of data;
ps5000getStream to get a packet of data;
ps5000close to close the communication.
In figure 7.1 the block diagram realized to setup the communication with the libraries. The
processing program is composed of the sub-routines described in chapter 4. Figure 7.2 shows the
block diagram of the core of the software with the call to the sub-routines. Finally, figure 7.3 is the
front panel of the program. In other words it is the graphical interface for the user, where it is
possible to setup all the parameters and the results (like the waterfall of the displacements) appear in
real-time.
Figure 7.1 The block diagram for the configuration of the streaming of data
106
Figure 7.2 The main block diagram
Figure 7.3 The front panel of the software
107
7.3 The differential sensor
Concerning the developments on the sensor components layout and design improvements, an
experimental campaign was focused on two main aspects. On one hand, it were studied the different
possibilities to detect and to measure the position and the displacement of an object (like a blade),
through different shielding materials (like Aluminium), without making holes on them. This is
possible thanks to the different magnetic permeabilities of the materials. For instance, the
magnetoresistive probes can detect the blade passage through an Aluminium layer, up to 10 mm
thickness with a SNR reduction of about 50%, but still good for the measurements. The advantage
is not only on avoiding to make holes in the casing to mount the sensors, but also on the
possibilities to use shielding material at the sensor’s top for applications in hot environments.
On the other hand, developments were focused on improving the repeatability when using different
materials. Passages of blades of different ferromagnetic materials at the same distance sensor-blade
tip, often results in different maximum voltage levels because of different polarization. This aspect
needs a normalization, then the calibration curve can be calculated. To avoid this, it was realized a
new configuration: the differential sensor of figure 7.4. Basically, two sensing elements are
mounted inside the sensor casing, shifted of a known quantity (i.e. 1 mm). In this way, at every
blade passage the probe generates two output signals, as reported in figure 7.5. The blue signal is
negated respect to the typical output because the sensor is mounted upsidedown. It is possible to
take the two absolute maximum peak values ( the dots in blue and green) then report in a graph
respect to the distance. At this point, there is only one linear curve passing through the two points
and this will be the calibration curve of the differential sensor (figure 7.6 left). Respect to the
normal sensor, the differential probe avoids the calibration procedure, at the cost of a less accurate
calibration curve because it is fitted on two points. Figure 7.6 right reports a comparison of the
calibration curves between a normal sensor calibrated in the range of [0.5 – 2.5] mm of blade tip
gap and the differential sensor mounted at 1 mm from the blade tip and its immediate calibration
curve extended to this range. The choice depends on several parameters, but where the accuracy do
not need to be very high, the differential sensor facilitates the classical procedures.
Figure 7.4 The differential magnetoresistive sensor
108
Figure 7.6 Comparison of the calibration curves
7.4 The Anti-Aliasing algorithm for tip-timing data
Finally, a new processing methodology to better analyze folded frequencies from BTT data was
studied. Figure 7.7 shows the measured displacements of the blade 1 of the SO-3 jet-engine
calculated with the algorithm developed in the frame of this project. By computing successive FFTs
with a window of 200 samples, it is possible to have the so-called Z-Plot of Figure 7.8. The blade
resonance around 9000 rpm appears in red. Moreover, there is a zig-zag patch that represents a
blade vibration mode presents up to about 12000 rpm. Its frequency is unknown because BTT data
is usually aliased due to the fact that it is possible to record only few blade vibration samples per
revolution and the Nyquist theorem is not satisfied. This results in a zig-zag patch. One way to
avoid the aliasing is to use several BTT sensors, which make the measurement chain expensive and
Figure 7.5 The two signals from the differential sensor
109
complex. A better way can be implemented from the software side, by considering the multiples of
the folded frequencies. An algorithm calculates the first N multiples of the measured folded
frequency (Figure 7.9) and selects among them the most probable true real frequency of vibration.
The selection procedure is a combination of methods, still under development. A preliminary result
on the data from the SO-3 Jet-Engine is shown in Figure 7.10. The ―cleaned‖ Zplot has only the
calculated real frequency of the vibration around 550 Hz, which is in good agreement with the FEM
Analysis of the Polish Institute.
Figure 7.7 Raw measured displacements calculated with the matlab software
Figure 7.8 The Zplot of an aliased frequency
110
Figure 7.9 The Zplot of the multiples of the aliased frequency
Figure 7.10 The Zplot after the anti-aliasing procedure
111
7.5 Conclusions
The realization of the differential sensor facilitates the calibration process and avoids the
normalization of the data explained in chapter 5.
A basic real-time measurement system was illustrated. More powerful hardware and more
sophisticated routines are needed to achieve the level of a professional multichannel BTT and BTC
system. Nevertheless, the one sensor real time system presented here, has given satisfactory results
in the calibration bench and it is an handy development platform for new applications.
112
8 Conclusions and Recommendations for Future Work
8.1 Conclusions
Various types of sensors have been proposed for Blade Tip Clearance and Blade Tip Timing
systems including electromagnetic, magnetic, inductive, optical, capacitance, eddy currents and
microwave. In this research project, a new sensor based on a magnetoresistive sensing elements has
been designed, realized and tested.
Clearance tests were carried out showing the possibility to perform very accurate measurements
with an uncertainty of U = ± 0,022 mm (20:1) within the range [1 ÷ 1.5] mm of gap variation, in
line with the current performance of commercial sensors. Tip-timing test cases were performed in
test benches and in real turbomachines with very satisfactory results. Two external collaborations
with Fogale Nanotech (France) and the ITWL (Poland) validated the system against different sensor
technology and has given the possibilities to improve further its performance.
Significant improvements were introduced during this project respect the first system configuration.
A differential magnetoresistive sensor was realized to avoid the calibration procedure and data
normalization. A real-time software and an innovative universal calibration bench for any BTT and
BTC sensor were developed. The principle of the blade motion decomposition can be now used for
simulation of different working conditions and development of new methods.
A database of technical information and a significant know-how on the use of magnetoresistive
sensors in turbomachinery testing is now available. The simultaneous BTC and BTT system
developed was successful in all test cases. It is still at the level of prototype, but it shows all the
requirements to be engineered at the level of a professional instrumentation.
8.2 Future Work
The future work should start from two main aspects:
First of all, the application of the magnetoresistive sensors should be studied for hot
environments. A new sensor configuration must be designed using a thermal shield or
innovative solutions. The first limitation in this kind of application will be the maximum
working temperature of the sensing element, which is currently limited at 125°C. Hence, a
thermal isolation system must be developed. From the side of the magnet, the limit is the
113
Curie temperature, which is much higher than 125°C. The application of an electromagnet to
substitute the permanent magnet should be considered. It also offers the possibility to tune
the magnetic field depending on the application.
Further developments are needed in terms of data processing methods. The anti-aliasing
software reported in chapter 7 should be improved and validated with strain gauge
measurements or similar. Innovative algorithms can decrease the amount of hardware
components needed for BTT and BTC measurements.
114
9 References
[1] T. C. Booth, Importance of tip clearance flows in turbine design. Tip Clearance Effects in Axial
Turbo-machines. Lecture Series, Von Karman Institute for Fluid Dynamics, 1985.
[2] D. Lombaert, Blade vibrations in turbomachinery -a literature study. Von Karman Institute for
Fluid Dynamics, U-VKI2008-10, 2007.
[3] K. Chana, Fundamentals of tip timing and tip clearance sensors. Tip timing and tip clearance
problems in turbomachinery. Lecture Series, Von Karman Institute for Fluid Dynamics, 2007.
[4] J. P. Roberts, The potentials of tip timing and tip clearance measurements for aero propultion
gas turbines, in Tip timing and tip clearance problems in turbomachinery, Lecture Series, Von
Karman Institute for Fluid Dynamics, 2007.
[5] A. von Flotow, M. Mercadal, and P. Tappert, Health monitoring and prognostics of blades and
disks with blade tip sensors, in Aerospace Conference Proceedings, 2000 IEEE, vol. 6, pp. 433^40,
IEEE, 2000.
[6] J. F. Brouckaert, Tip timing and tip clearance problems in turbomachinery. Lecture Series, Von
Karman Institute for Fluid Dynamics, 2007.
[7] D. Cardwell, K. Chana, and P. Russhard, The use of eddy current sensors for the measurement
of rotor blade tip timing-sensor development and engine testing, ASME GT2008-50792, 2008.
[8] J. Gallego-Garrido, G. Dimitriadis, and J. R. Wright, A class of methods for the analysis of
blade tip timing data from bladed assemblies undergoing simultaneous resonances part ii:
Experimental validation, International Journal of Rotating Machinery, vol. 2007, 2007.
[9] S. Heath and M. Imregun, An improved single-parameter tip-timing method for
turbomachinery blade vibration measurements using optical laser probes, International journal of
mechanical sciences, vol. 38, no. 10, pp. 1047-1058, 1996.
[10] H. S. Dhadwal and A. P. Kurkov, Dual-laser probe measurement of blade-tip clearance,
Journal of turbomachinery, vol. 121, Issue 3, pag. 481-485, 1999.
115
[11] T. Pfister, L. Buttner, J. Czarske, H. Krain, and R. Schodl, Turbomachine tip clearance
and vibration measurements using a fibre optic laser doppler position sensor, Measurement
Science and Technology, vol. 17, no. 7, p. 1693, 2006.
[12] M. Zielinski and G. Ziller, Noncontact blade vibration measurement system for aero engine
application, in 17th International Symposium onAirbreath-ing Engines, Paper No. ISABE-2005-
1220, 2005.
[13] S. Lavagnoli, G. Paniagua, M. Tulkens, and A. Steiner, High-fidelity rotor gap
measurements in a short-duration turbine rig, Mechanical Systems and Signal Processing, vol. 27,
pp. 590-603, 2011.
[14] G. W. Terpay, Shielded Eddy Current Sensor For Enhanced Sensitivity, United states patent
5,942,893, 1999.
[15] R. Grzybowski, G. Foyt, W. Atkinson, W. Knoell, J. Wenger, and D. Benz, Microwave blade
tip clearance measurement system. ASME96-GT-2, 1996.
[16] A. Maslovskiy, Microwave Turbine Tip Clearance Measuring System for Gas Turbine Engines.
ASME-GT2008-50354, 2008.
[17] M. Wagner, A. Schulze, M. Vossick, C. Stephelbauer, R. Weigel, N. Vortmeyer, and P. Heide,
Novel microwave vibration monitoring system for industrial power generating turbines, in
Microwave Symposium Digest, 1998 IEEE MTT-S International, vol. 3, pp. 1211-1214, IEEE,
1998.
[18] C. Belsterling, Sensing with air, Instrumentation Science and Technology, vol. 18, no. 3, pp.
37—41, 1971.
[19] J. Rickman Jr, Eddy current turbocharger blade speed detection, IEEE Transactions on
Magnetics, vol. 18, no. 5, pp. 1014-1021, 1982.
[20] C. Roeseler, A. von Flotow, and P. Tappert, Monitoring blade passage in turbomachinery
through the engine case (no holes), in Aerospace Conference Proceedings, 2002. IEEE, vol. 6, pp.
6-3125, IEEE, 2002.
[21] J. Gallego-Garrido, G. Dimitriadis, and J. Cooper, Further Developments of Blade Tip-timing
Data Analysis Techniques. PhD thesis, 2005.
116
[22] I. B. Carrington, Development of blade tip timing data analysis techniques. PhD thesis,
University of Manchester, 2002.
[23] G. Rossi, Progettazione, realizzazione e caratterizzazione di un sistema di misura di vibrazioni
di organi rotanti di turbomacchine con sensori stazionari. PhD thesis, Facolta di ingegneria
Universita di Ancona, 1993.
[24] I. Y. Zablotskiy and Y. A. Korostelev, Measurement of resonance vibrations of turbine blades
with the elura device, Energomashinostroneniye,(USSR), no. 2, pp. 36-39, 1970.
[25] S. Heath, T. Slater, L. Mansfield, and P. Lof-tus, Turbomachinery blade tip-timing
measurement techniques, Proceedings of Advisory Group for Aerospace Research and Development
Conference, 90th Symposium on Advanced Non Intrusive Instrumentation for Propulsion Engines,
pp. 32-1 32-9, Brussels, Belgium, 1997.
[26] S. Heath and M. Imregun, A survey of blade tip-timing measurement techniques for
turbomachinery vibration, Journal of engineering for gas turbines and power, vol. 120, no. 4, pp.
784-791, 1998.
[27] M. Zielinski and G. Ziller, Noncontact vibration measurements on compressor rotor blades,
Measurement Science and Technology, vol. 11, no. 7, p. 847, 2000.
[28] G. Dimitriadis, I. B. Carrington, J. R. Wright, and J. E. Cooper, Blade-tip timing
measurement of synchronous vibrations of rotating bladed assemblies, Mechanical Systems and
Signal Processing, vol. 16, n. 4, pp. 599-622, 2002.
[29] I. B. Carrington, J. R. Wright, J. E. Cooper, and G. Dimitriadis, A comparison of blade tip
timing data analysis methods, Proceedings of the Institution of Mechanical Engineers, Part G:
Journal of Aerospace Engineering, vol. 215, no. 5, pp. 301-312, 2001.
[30] J. Gallego Garrido, G. Dimitriadis, and J. R. Wright, Development of a multiple modes simulator
of rotating bladed assemblies for blade tip timing data analysis, in Proceedings of the 27th
International Conference on Noise and Vibration Engineering ISMA02, pp. 1437-1446, 2002.
[31] J. Gallego Garrido and G. Dimitriadis, Validating synchronous blade vibration amplitudes
from blade tip-timing data analysis, Institution of Mechanical Engineers, vol. 623, no. 23, pp. 205-
213, 2004.
117
[32] E. Armstrong and R. Stevenson, Some practical aspects of compressor blade vibration, The
Journal of the Royal Aeronautical Society, vol. 64, no. 591, pp. 117-130, 1960.
[33] W. Robinson, Overview of pratt and whitney nsms, in 45th International Instrumentation
Symposium, Albuquerque, New Mexico, 1999.
[34] W. Robinson and R. Washburn, A real time noninterference stress measurement system (nsms)
for determining aero engine blade stresses, in Proceedings of the 37th Symposium of the
Instrumentation Society of America, San Diego, California, pp. 91-103, 1991.
[35] H. T. Jones, Development of advanced non-intrusive stress measurement processing techniques,
in Proceedings of the 6th National Turbine Engine High Cycle Fatigue Conference, Jacksonville,
Fla, USA, 2001.
[36] B. W. Hayes, S. Arnold, C. Vining, and R. Howard, Application of generation 4 non-contact
stress measurement system on hcf demonstrator engines, in Technical Papers of ISA, vol. 451, pp.
147-163, 2004.
[37] J. Gallego-Garrido, G. Dimitriadis, and J. R. Wright, A class of methods for the analysis of
blade tip timing data from bladed assemblies undergoing simultaneous resonances part i:
Theoretical development, International Journal of Rotating Machinery, vol. 2007, 2007.
[38] Ewins, D. (1992b). Structural dynamic characteristics of individual blades, in Vibration and
rotor dynamic, Lecture Series 1992-06. VKI.
[39] Rao, J. (1991). Turbomachine blade vibration. New York: Wiley.
[40] Haupt, U. (1987a). Blade vibration problems in centrifugal compressors, in Flow in centrifugal
compressors, Lecture Series 1987-01 (Vol. A). VKI.
[41] Hasemann, H., Weser, G., Hagelstein, D., & Rautenberg, M. (1997). Investigation of the
solidity and the blade vibration behaviour of unshrouded centrifugal compressor impellers with
different aerodynamic design, ASME, Vol. 97GT233.
[42] Rao, S. S. (2007). Vibration of continuous systems, Hoboken: Wiley.
[43] Imregun, M. (1999). Aeroelasticity in axial Turbomachine, in , Lecture Series 1999-05. VKI.
[44] Ewins, D. (1992a). Structural dynamic characteristics of bladed disks, in Vibration and rotor
dynamic, Lecture Series 1992-06. VKI.
118
[45] Singh, M. P., Vargo, J. J., Schiffer, D. M., & Dello, J. D.(2000). SAFE Diagram, A design and
reliability tool for turbine blading, Wellsville, New York: Dresser-Rand.
[46] Mehdigholi, H. (1991). Forced Vibration of Rotating Disc and Interaction with Non-rotaring
structures. Ph.D. Thesis, College of Science, Technology andMedicine, London.
[47] JAY, R., MACBAIN, J., & BURNS, D. (1984). Structural response due to blade vane
interaction. Journal of Engineering for Gas Turbines and Power, Vol.106, No 1, 50-56.
[48] Kushner, F. (2000). Compressor blade and impeller rotating disk vibration avoidance
parameters. The American Society of Mechanical Engineers.
[49] Judge, J., Pierre, C., & Mehmed, O.(2001). Experimental investigation of mode localization
and forced response amplitude magnification for a mistuned bladed disk. Journal of Engineering for
Gas Turbines and Power, Vol.123, No 4, 940-950.
[50] Seinturier, E. (2005b). Forced response computation for bladed disks industrial practices and
advanced methods, in Effects of aerodynamic unsteadiness in axial turbomachines, Lecture Series
2005-03. VKI.
[51] Singh, M., & Vargo, J.(1989). Reliability evaluation of shrouded blading using the SAFE
interference diagram. Journal of Engineering for Gas Turbines and Power, Vol.111, No 4, 601-609.
[52] J. D. Denton. Loss Mechanisms in Turbomachines. Journal of Turbo-machinery, 115(4), pp
621-656, 1993.
[53] B. Glezer. Transient Tip Clearance Analysis and Measurement Techniques. Number 02 in
Turbine Blade Tip Design and Tip Clearance Treatment. Von Karman Institute for Fluid Dynamics,
Rhode Saint Genese, Belgium, January 2004.
[54] P.E.McCarty, W.Thompson Jr., Development of non interference technique for measurement
of turbine engine compressor blade stress, Arnold Eng. Development Center Technical Report, 79-
78, USAir Force, Tennessee, USA
[55] L.J.Kiraly, Digital system for dynamic turbine engine blade displacement measurements,
NASA TM - 81382.
[56] H.Roth, Vibration and clearance measurements on rotating blades using stationary probes,
Von Karman Institute [Lecture Series 81-7], (1981), Rhode Saint Genese, Belgium.
119
[57] Rossi G., Brouckaert J.-F., Design of a blade tip timing measurement system based on
uncertainty analysis, 58th
ISA International Instrumentation Symposium, San Diego (CA), June 4-7,
2012.
[58] P. A. Holman, Magnetoresistance (MR) Transducers And How to Use Them as Sensors, 2004.
[59] http://www.magneticsensors.com/literature.php
[60] Honeywell, Magnetic Displacement Sensors HMC1501/1512, (datasheet)
[61] Procházka, P. and F. Vaněk. (2010).Contactless Diagnostics of Blade Vibration of Low-
Pressure Steam Turbine Stages, Proceedings of ISMA2010 International Conference on Noise and
Vibration Engineering including USD2010. Leuven,Vol. 2420.
[62] E. Cardelli, A. Faba, R. Marsili, G. Rossi, and R. Tomassini, Magnetic nondestructive testing
of rotor blade tips, Journal of Applied Physics 117.17, 17A705 (2015); doi: 10.1063/1.4907180.
[63] E. Cardelli, A. Faba, F. Tissi, Surface field measurements in vector characterization of Si-Fe
magnetic steel samples, International Journal of Applied Electromagnetics and Mechanics, Volume
44, Number 3-4, pages 331-338, April 2014.
[64] E. Cardelli, A. Faba, Vector hysteresis measurements of not oriented grain SiFe steels by a
biaxial hall sensors array, Physica B: Condensed Matter, Vol. 435, p. 34–39, February 2014.
[65] L. Andrenelli, N. Paone and G. L. Rossi, Large-bandwidth reflection fiber-optic sensors for
turbomachinery rotor blade diagnostics, in Sensors and Actuators A, vol 32, p. 539-542, 1992.
[66] P. Nava, N. Paone, G.L. Rossi and E.P. Tomasini, Design and experimental characterization of
a nonintrusive measurement system of rotating blade vibration, Journal of engineering for gas
turbines and power, vol. 116, p. 657-662, 1994.
[67] J.-F. Brouckaert, R. Marsili , G. Rossi, and R. Tomassini, Development and
experimental characterization of a new non-contact sensor for blade tip timing, in AIP Conference
Proceedings, vol. 1457, p. 61, 2012.
[68] R. Tomassini, G. Rossi, J. Brouckaert, On the development of a magnetoresistive sensor for
blade tip timing and blade tip clearance measurement systems, 11th International conference on
vibration measurements by laser and noncontact techniques AIVELA 2014: Advances and
Applications, Vol.1600, AIP Publishing, 2014, pp. 65–73.
120
[69] Diek, R. H., (2007), Measurement Uncertainty Methods and Applications, Systems and
Automation Society (ISA)
[70] Castrup, S., and Castrup, HT, (2010), Measurement Uncertainty Analysis Principles and
Methods, National Aeronautics and Space Administration
[71] Lattime, S.B., Bruce, M.S., (2004), High-pressure-turbine clearance control systems: current
practices and future directions, Journal of Propulsion and Power, 20.2, 302-311.
[72] Przysowa R. (2014) Analysis of synchronous blade vibration with the use of linear sine fitting,
Journal of KONBiN, 2(30), pp. 5-20, De Gruyter Open.
121
Acknowledgements
Thanks go to Professor Gianluca Rossi , Professor Jean-Francois Brouckaert and Professor Tony
Arts for the continual advices, support and encouragements provided throughout all the course of
this research.
I wish to express my gratitude to Professor Gianluca Rossi and the University of Perugia, to
Professor Giampiero Naletto and the CISAS as well as the von Karman Institute to have made this
research project possible.
The author would like to acknowledge all the von Karman Institute family for the wonderful
international experience, the atmosphere and the personal growth.
The experimental campaign would not have been possible without the precious help and technical
advices of Jean-Christophe Desilve, Terence Boyen, Pierre Cuvelier, as well as all the technician
staff. Particular thanks to Christophe and Pierre, to have always listened the second question
coming after ―just one question!‖
For the motivation, the continuous encouragements and to have always tried to increase my self-
confidence, a special thank goes to Susanna.
A big thank you goes to all that friends that were with me, despite the distance. Particular thanks to
Enrico, Matteo, Maurizio, Nicola and Alex.
Finally, I would like to thank my family: Angiolo, Anna Maria and Matteo, for their continuous
support and encouragement throughout all my life and Paulina, to have entered in my life.
122
Appendix
A. The VKI R2 test rig
The VKI high speed compressor facility R-2 is a semi-closed loop test rig for investigation of axial
and radial compressors (Fig. 1.). The axial compressor can have a maximum diameter of 500 mm
and the casing is large enough to house a two-stage configuration. The facility is driven by a 185
kW DC motor. The rotational speed can vary up to 10,000 RPM and is monitored by an electronic
speed control.
Figure A1 The VKI compressor facility R-2.
The current configuration is a single stage axial compressor installed in a 400 mm diameter test
section. The compressor inlet is composed of a bell-shaped inlet duct as shown in Figure 1. The
rotor (Fig. 2) is composed of 24 subsonic blades of the NACA 65 family with a hub radius of 100
mm and a tip radius of 200 mm. Figure 4 represents a characteristic section of the rotor blade. The
main geometrical characteristics are presented in Table 1. Little is known about the aerodynamic
design of this machine because the facility was entirely built around the rotor which was a donation
from DLR to VKI in the early 70’s for educational purposes.
The stator (Fig. 3) is composed of 30 blades of 50 mm chord made of simple bent and twisted thin
metal plates to remove the swirl from the stage exit flow before entering the collector. The stator
leading edge plane is located 85 mm downstream of the rotor trailing edge plane (at midspan). After
123
leaving the stator, the flow is discharged to atmosphere through a collector and an air cooler. The
throttling of the massflow is obtained by closing a perforated drum in the collector downstream of
the compressor. The valve rotation is tuned using a small 24 V DC electric motor.
Figure A2 The R-2 compressor rotor. Figure A3 The R-2 compressor stator.
Figure A4 The R-2 compressor rotor blade (NACA 65 type).
124
Table A1 Blade geometry of the R-2 compressor rotor.
The performance map of the compressor is presented in Figure 5 where the nominal rotational
speed corresponds to 10,000 RPM. Presently, the maximum rotational speed of the compressor is
limited to 6000 RPM. This is due to the limitation on peripheral speed of the seals used on the shaft.
Figure A5 Performance map of the R-2 compressor.
125
Instrumentation
Figure 6 shows a general view of the facility. Four circumferential locations (every 90º) allow the
insertion of probes upstream and downstream of the rotor.
Figure A6 Instrumentation on the VKI high speed compressor facility R-2.
126
Figure 7 shows a view of one of the 4 instrumentation access windows (40mm x 285mm) showing
also the rotor blade tip. The angular sector corresponding to the opening is 11.65 degrees. The
access window is closed by an aluminium plate which allows easy modifications for insertion of
probes, pressure sensors, tip timing and tip clearance sensors, infinite line probes, etc. A technical
sketch of such a plate is given in Figure 8. A standard adaptor fitting could be specified in common
agreement with sensor providers to fit in different sensors at the same location.
Figure A7 View of an instrumentation access window.
Figure A8 Sketch of instrumentation closing plate.
127
Other specific aspects of the rig related to tip timing
An existing inlet distortion grid may be installed in front of the rotor. It is composed of 4 sectors
which can be installed together or separately at various axial distances from the rotor leading edge.
The grid is made of 24 cylindrical bars of 3 mm diameter. Figure 9 shows a quarter of the inlet
distortion grid installed in front of the rotor.
Figure A9 Partial inlet distortion grid installed at the compressor inlet.
Regarding tip clearance measurements, simple thickness gauge data have been taken at leading
edge, mid-chord and trailing edge of the rotor, which are shown in Fig. 10. The average tip
clearance is ~ 0.64 mm.
Figure A10 Rotor tip clearance distribution.
128
B. The magnetoresistive sensing element - Honeywell HMC1501
Figure B1 Schematic diagrams and package outline
129
Figure B2 Specifications
Figure B3 Basic electronic circuit
130
Figure B4 Scheme of the electric circuit of the analogue sensor
131
Figure B5 Scheme of the electric circuit of the digital sensor
132
C. The data acquisition system
Figure C1 The PicoScope 4424
Figure C2 The PicoScope 5204
133
D. Blade Tip Clearance
Figure D1 Acquired signal at 1000rpm
Figure D2 Acquired signal at 3000rpm
134
Figure D3 Acquired signal at 5000rpm
Uncertainty analysis
Table D1 Random and Systematic components of the uncertainty
135
Considering the above mentioned quantities, in this case the values that have been taken during the
calibrations, are according to what follows:
-The systematic component has been considered equal to the comparator resolution = 0.01mm;
-The standard deviation has been evaluated during the fitting of the experimental data. N, the
number of points used to obtain the calibration curve was 23 (Figure 2.15).
systematic component standard deviation
N random
component
calibration 0,01 0,00244875 23 0,0005106
t95 2,069
U95 0,021878087
Table D2 Uncertainty calculations
The estimated value of uncertainty is reported in Table D.2. The definition used for the calculation
is shown below, with the values of the t-student’s distribution.
Assumption:
At that stage, the temperature effect has not been taken into account. This effect has been
considered negligible considering that, the test duration was less than 60 seconds, performed at
ambient temperature, in the lathe, at 45 rpm. Moreover, the sensing element is a Wheatstone bridge
of Permalloy elements and it should compensate the temperature unbalance itself. Anyway, for
future characterization at high speed, it should be considered.
136
E. The Universal Calibration Bench
Figure E1 Maximum achievable displacements in function of the frequency
Figure E2 The new rotor blade
137
Figure E3 The new rotor disk
138
F. Optimization of the sensor components
The magnets used during the experimental investigation have been bought form:
http://www.supermagnete.de/eng
Figure F1 The cylindrical magnet
Figure F2 The cubic magnet
139
Figure F3 The linear slider
140
Figure F4 Base plate drawings
141
Figure F5 The drawing of the blade
Figure F6 The shaft and the flange
142
Figure F7 The sensor housing support
Figure F8 Rendering of the CAD assembly
143
y axis
position [mm]
20 15 10 5 0
slider [mm]
0 5 10 15 20
x ax
is
39 -10 39,50 39,25 35,32 32,25 30,28
34 -5 39,80 39,70 37,15 34,02 32,30
29 0 37,24 37,03 35,82 34,19 33,45
24 5 20,08 28,43 30,33 31,08 31,75
19 10 14,34 14,60 20,69 24,77 27,00
14 15 23,75 8,80 9,38 16,50 19,95
9 20 34,95 22,42 5,09 7,76 11,02
Table F1 Peak to peak output signal values
Figure F9 The different output signal shapes observed
144
Figure F10 The maximum output signals obtained with the cylindrical magnet (top) and with the cubic magnet (bottom)